No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Abstract
This paper proposes genetic algorithms (GAs) approach to feature discretization and the determination of connection weights for artificial neural networks (ANNs) to predict the stock price index. Previous research proposed many hybrid models of ANN and GA for the method of training the network, feature subset selection, and topology optimization. In most of these studies, however, GA is only used to improve the learning algorithm itself. In this study, GA is employed not only to improve the learning algorithm, but also to reduce the complexity in feature space. GA optimizes simultaneously the connection weights between layers and the thresholds for feature discretization. The genetically evolved weights mitigate the well-known limitations of the gradient descent algorithm. In addition, globally searched feature discretization reduces the dimensionality of the feature space and eliminates irrelevant factors. Experimental results show that GA approach to the feature discretization model outperforms the other two conventional models.
... The corresponding dynamic hedginginspired controller shall be referred to as SMPC-DH. Both the difference with respect to dynamic option hedging and the motivation for employing (8) for stock trading are visualized in Figure 2. Using (8) in combination with (7) can be interpreted as a trailing stop-loss strategy. Finally, we remark that the other two stochastic measures (QP-Var and LP-CVaR) from [4] can be employed likewise using (8). ...
... The corresponding dynamic hedginginspired controller shall be referred to as SMPC-DH. Both the difference with respect to dynamic option hedging and the motivation for employing (8) for stock trading are visualized in Figure 2. Using (8) in combination with (7) can be interpreted as a trailing stop-loss strategy. Finally, we remark that the other two stochastic measures (QP-Var and LP-CVaR) from [4] can be employed likewise using (8). ...
... Both the difference with respect to dynamic option hedging and the motivation for employing (8) for stock trading are visualized in Figure 2. Using (8) in combination with (7) can be interpreted as a trailing stop-loss strategy. Finally, we remark that the other two stochastic measures (QP-Var and LP-CVaR) from [4] can be employed likewise using (8). ...
We seek a discussion about the most suitable feedback control structure for stock trading under the consideration of proportional transaction costs. Suitability refers to robustness and performance capability. Both are tested by considering different one-step ahead prediction qualities, including the ideal case, correct prediction of the direction of change in daily stock prices and the worst-case. Feedback control structures are partitioned into two general classes: stochastic model predictive control (SMPC) and genetic. For the former class three controllers are discussed, whereby it is distinguished between two Markowitz- and one dynamic hedging-inspired SMPC formulation. For the latter class five trading algorithms are disucssed, whereby it is distinguished between two different moving average (MA) based, two trading range (TR) based, and one strategy based on historical optimal (HistOpt) trajectories. This paper also gives a preliminary discussion about how modified dynamic hedging-inspired SMPC formulations may serve as alternatives to Markowitz portfolio optimization. The combinations of all of the eight controllers with five different one-step ahead prediction methods are backtested for daily trading of the 30 components of the German stock market index DAX for the time period between November 27, 2015 and November 25, 2016.
... • The application of evolutionary algorithms for feature discretization in neural networks for stock price prediction was investigated by Kim and Han (2000). They illustrated the significance of financial measures of stock performance, such as profit and sales [1]. ...
... • The application of evolutionary algorithms for feature discretization in neural networks for stock price prediction was investigated by Kim and Han (2000). They illustrated the significance of financial measures of stock performance, such as profit and sales [1]. • Support Vector Machines, a reliable classifier that has shown efficacy in high-dimensional and non-linear datasets, such those found in stock forecasting, were first presented by Cortes and Vapnik in 1995 [2]. ...
This study uses Support Vector Machine (SVM) models to estimate stock prices by examining the link between a company's sales, profit, and future orders. Finding out how well these financial indicators can forecast changes in stock prices is the goal. To forecast the future direction of stock prices, we use past financial data from publicly traded companies to construct an SVM model. With future orders exhibiting the highest link, the findings reveal that these financial indicators play a substantial role in predicting the direction of stock prices.
... In addition, models built upon extremely versatile tree models are also popular, such as decision tree [19], [20], [21], random forest [22], gradient boosting decision tree [23]. Apart from that, Artificial Neural Networks (ANN) [24], [25], [26], Genetic algorithms (GA) [24], [27], [28], and the Hidden Markov Model (HMM) [29] are three methods used to forecast the behavior of the financial markets. Moreover, Support Vector Machines (SVMs) have shown better performance in forecasting stock market changes [30]. ...
... In addition, models built upon extremely versatile tree models are also popular, such as decision tree [19], [20], [21], random forest [22], gradient boosting decision tree [23]. Apart from that, Artificial Neural Networks (ANN) [24], [25], [26], Genetic algorithms (GA) [24], [27], [28], and the Hidden Markov Model (HMM) [29] are three methods used to forecast the behavior of the financial markets. Moreover, Support Vector Machines (SVMs) have shown better performance in forecasting stock market changes [30]. ...
The endeavor of stock trend forecasting is principally focused on predicting the future trajectory of the stock market, utilizing either manual or technical methodologies to optimize profitability. Recent advancements in machine learning technologies have showcased their efficacy in discerning authentic profit signals within the realm of stock trend forecasting, predominantly employing temporal data derived from historical stock price patterns. Nevertheless, the inherently volatile and dynamic characteristics of the stock market render the learning and capture of multi-scale temporal dependencies and stable trading opportunities a formidable challenge. This predicament is primarily attributed to the difficulty in distinguishing real profit signal patterns amidst a plethora of mixed, noisy data. In response to these complexities, we propose a Multi-Scale Temporal Memory Learning and Efficient Debiasing (MTMD) model. This innovative approach encompasses the creation of a learnable embedding coupled with external attention, serving as a memory module through self-similarity. It aims to mitigate noise interference and bolster temporal consistency within the model. The MTMD model adeptly amalgamates comprehensive local data at each timestamp while concurrently focusing on salient historical patterns on a global scale. Furthermore, the incorporation of a graph network, tailored to assimilate global and local information, facilitates the adaptive fusion of heterogeneous multi-scale data. Rigorous ablation studies and experimental evaluations affirm that the MTMD model surpasses contemporary state-of-the-art methodologies by a substantial margin in benchmark datasets. The source code can be found at https://github.com/MingjieWang0606/MDMT-Public.
... In order to find new opportunities in the field of optimality, many researchers have contributed their works in various application domains. For example, the article in [20] proposes a GA approach to determine the weights of the interconnected nodes for the ANN model from the perspective of the economics domain. Similarly, the weights of another ANN model are determined in [21] by using GA. ...
... Similarly, the weights of another ANN model are determined in [21] by using GA. However, the tuning of hyper-parameters to determine the suitable neural architecture is beyond the scope of the work presented in [20,21]. Furthermore, the combined features of GA and ANN are emphasized to quantitatively assess the methodology used in [22]. ...
A major impact on energy savings by efficient use of streetlights becomes significant to develop a smart city. There is a need to optimize the values of necessary street lighting parameters for energy planning. So, an artificial neural network embedded genetic algorithm approach is proposed to obtain energy efficient street lighting. The neural network serves the purpose of the autonomous dimming of streetlights whenever required, while the genetic algorithm is utilized to minimize the error anticipating in the training procedure of the artificial neural network model. In the proposed work, each individual population of the genetic algorithm is comprised of four genes, such as the number of hidden layers as well as the number of neurons on the first layer in the artificial neural network, the activation function and the optimizer used for the training of the artificial neural network model. The outcome of the genetic algorithm generates the suitable values of the hyper-parameters in combinations, which in turn determines the minimum training error for the artificial neural network model. The effectiveness of tuning hyper-parameters for selecting the best neural network architecture is comprehensively assessed. The computation time of the proposed work for different variations in terms of optimizers and activation functions is shown. Finally, the proposed embedded framework has shown an improvement of 41.94% more energy efficiency and 94.8% less training error, instead of existing works.
... There are various accuracy measures available, such as absolute forecasting errors, percentage errors, symmetric errors, measures based on relative errors, and scaled and relative measures, but directional accuracy measures assume higher importance in stock market prediction (Majumder & Hussain, 2010). Accuracy rate, a directional measure, has been preferred by various researchers for predicting the accuracy of models (Kim & Han, 2000;Kim, 2003;Choudhary & Garg, 2008;Cao & Tay, 2003;Altay & Satman, 2005;Chun & Park, 2005;Gestel et al., 2001;Huang et al., 2005;Roh, 2007;Siekmann et al., 1999;Kara et al., 2011;Ou & Wang, 2009;Yu et al., 2009;Kumar & Thenmozhi, 2006;Kim, 2006;Mizuno et al., 1998;Yao & Poh, 1995;Wunsch et al., 1998). ...
... Given the nature of stock market predictions, it has become necessary to move from simple mathematical model accuracy to model profitability. On account of the influence and guidance of physical sciences and engineering in the context of machine learning techniques, the majority of the previous studies ignored profitability-based assessment of models (Altay & Satman, 2005, Choudhry & Garg, 2008, Gestel et al., 2001, Chun & Park, 2005, Siekmann et al., 1999, Kara et al., 2011, Ou & Wang, 2009, Yu et al., 2009, Kim, 2006, Kim, 2003, Kim & Han 2000. The evaluation of predictive models is a crucial dimension in context of stock market prediction. ...
Understanding the linkage between directional accuracy and return remains a negligibly explored domain in the context of stock market prediction. Maximum studies in the given domain are based on mathematical accuracy measures and directional accuracy measures. Merely the efficiency of directional prediction may not suffice, as themagnitude of directional change also assumes importance in stock markets. Without any doubt, profitability measures reign supreme in the case of the stock market. The current study attempts to investigate the association between Accuracy rate and returns for data sets based on seven stock indices across the globe. The outcome of the study relies on machine learning: ANN (Artificial Neural Networks) and SVM (Support Vector Machine) based predictive models, built using 196 data sets of variable duration. The results of the current research reveal a strong positive correlation between accuracy rate and returns. Moreover, significant variations in the extent of therelationship are also observed. Quantile regression models indicate a relatively stronger relationship between accuracy rate and return at the lower range of returns. Findings of the study carry value for academicians, researchers, and practitioners in the field of machine learning and stock market predictions.
... Evolutionary algorithms have been widely applied to neural networks, and various hybrid approaches have also been used for economic time series prediction. Genetic algorithms allow the learning algorithm to be improved by acting as a network training method, feature subset selection, and neural network topology optimization, as well as reducing the complexity of the feature space [34]. Figure 3 shows the hybrid GA-LSTM approach to find the time window size and the number of LSTM units for the prediction of the unemployment rate in Ecuador. ...
... In the case of the individual algorithms, combinations were tested using the Grid-SearchCV technique present in the Scikit-learn library: the number of hidden layers is (1, 2), the number of hidden units is (2,3,4,5,6,7,8), the batch number is (2,4,8,10,12,14), the number of epochs is (32,34,36,60,65,70,75,80), and the optimizers are (RMSprop, Adam). ...
Unemployment, a significant economic and social challenge, triggers repercussions that affect individual workers and companies, generating a national economic impact. Forecasting the unemployment rate becomes essential for policymakers, allowing them to make short-term estimates, assess economic health, and make informed monetary policy decisions. This paper proposes the innovative GA-LSTM method, which fuses an LSTM neural network with a genetic algorithm to address challenges in unemployment prediction. Effective parameter determination in recurrent neural networks is crucial and a well-known challenge. The research uses the LSTM neural network to overcome complexities and nonlinearities in unemployment predictions, complementing it with a genetic algorithm to optimize the parameters. The central objective is to evaluate recurrent neural network models by comparing them with GA-LSTM to identify the most appropriate model for predicting unemployment in Ecuador using monthly data collected by various organizations. The results demonstrate that the hybrid GA-LSTM model outperforms traditional approaches, such as BiLSTM and GRU, on various performance metrics. This finding suggests that the combination of the predictive power of LSTM with the optimization capacity of the genetic algorithm offers a robust and effective solution to address the complexity of predicting unemployment in Ecuador.
... However, ANN models often suffer from overfitting, local optima issues, and sensitivity to financial data noise. RNNs have also been explored but face gradient vanishing and exploding problems [5]. To overcome these limitations, LSTM networks have been adopted, offering improved performance in capturing long-term dependencies [6]. ...
Accurate time series forecasting is essential for decision-making in financial markets, power generation, and economic planning. However, traditional models often struggle to capture the complex nonlinear patterns in financial data, leading to suboptimal predictions. To address this, we propose a novel hybrid approach integrating the Whale Optimization Algorithm (WOA) with Support Vector Machines (SVM) for enhanced stock price forecasting. The WOA-SVM model optimizes key SVM hyperparameters—Regularization Parameter (????) and Kernel Coefficient (????)—while also performing feature selection to improve model generalization. By effectively balancing exploration and exploitation, WOA accelerates convergence, reduces computational complexity, and minimizes forecasting errors. Extensive experiments on S&P 500 and NIFTY 50 datasets confirm WOA-SVM’s superiority over SVM, LSTM, and Random Forest Regression, achieving the lowest MSE (2.45) and RMSE (1.56). These results highlight WOA-SVM as a robust and efficient tool for financial market forecasting, offering valuable insights to investors, analysts, and financial institutions.
... Recent developments build on this basis by using advanced methods including clever approaches to improve prediction precision. For stock price index predictions, Kim and Han, for example, used evolutionary algorithms housed within artificial neural networks [3]. Chen and Lin paired feature selection techniques [4] with Support Vector Machines (SVM). ...
In the highly competitive and constantly evolving landscape of the investment realm, the construction of an optimal investment portfolio has emerged as a cornerstone for investors aiming to adeptly navigate the complex interplay between risks and returns. With the rapid advancement of technology, the performance and trends of stocks have become increasingly intertwined with technological developments. This study analyzes five tech-dependent stocks and applies the Long Short-Term Memory (LSTM) model for price prediction. The combination of LSTM and Mean-Variance Optimization (MVO) model outputs daily portfolio weights using two strategies: minimizing variance and maximizing Sharpe ratio. The minimal variance model produces a cumulative return of 193.1164% during a 30-day testing period from day 71 to day 100; the maximum Sharpe ratio model produces a return of 61.2246%, both exceeding the return of the CSI 300 Index at just 14.6193%. The outcomes validate the effectiveness of the suggested approach and provide insightful analysis for portfolio managers choosing their course of action.
... YSA yöntemiyle, Leigh vd. (2002) New York Menkul Kıymetler Borsası Endeksi'ni; Kim ve Han (2000) borsa endeksinin gelecekte alacağı değeri tahmin etmeye çalışmışlardır. Dhar ve Chou (200 ı) firmaların gelecekte elde edecekleri kazançların ve getirilerin tahmininde kullanılan doğrusalolmayan yöntemleri karşılaştırmışlar ve daha önce yapılmış olan çalışmalann doğrusal istatistik yaklaşımları içerdiğini, son yıllarda yapılan çalışmalarda ise doğrusalolmayan yöntemlerin benimsendiğini vurgulamışlardır. ...
Çalışmanın amacı işletme problemlerine uygulanabilecek yöntemler arasında yaygınlaşmakta olan yapay sinir ağlarını (YSA) tanıtmak, finans alanında uygulanabilirliğini araştırıp Türkiye ile ilgili bir uygulama yaparak katkıda bulunmaktır. YSA, verideki ilişki kalıplarını ve eğilimleri tanıyarak öngörüm veya veri sınıflaması yapar. YSA, problemi modellemeye ve istatistiksel varsayımlar yapmaya gerek duymadan örneklerle öğrenir. Ancak optimal ağ topolojisini belirleyen bir yöntem yoktur. Ampirik çalışmada, YSA kullanarak yedi şirketin hisse senedi fiyatı tahmin edilmiştir. İlk aşamada YSA yöntemi haftalık ve günlük verilere uygulanmış, günlük veri kullanımı daha başarılı sonuçlar vermiştir. İkinci aşamada günlük verilerle YSA ve doğrusal regrasyon yöntemleri denenmiştir. YSA’nın regrasyondan daha başarılı sonuçlar vermesi, alternatif finansal analiz yöntemleri arasında değerlendirilebileceğini gösterir.
... Many different methods are used in forecasting financial markets. These include ARIMA (Box et al., 2015;Kulkarni et al., 2020), GARCH (Gabriel & Ugochukwu, 2012, Chen & Chen, 2015, genetic algorithm (Nikolopoulos & Fellrath, 1994;Kim & Han, 2000), ANN (Roh, 2007;Vijaya et al., 2016;Gurjar et al., 2018;Gaytan et al., 2022;Kurani, 2023); machine learning algorithms (Umer et al., 2019;Khan et al., 2020;Vijh et al., 2020;Rouf et al., 2021;Soni et al., 2022;Kumbure et al., 2022;Ashtiani & Raahmei, 2023;Jorgenson et al., 2023), deep learning algorithms (Jiang, 2021;Mehtab et al., 2021;Hu et al., 2021;Li&Pan, 2022;Shah et al., 2023;Muhammad et al., 2023;Mukherjee et al., 2023) methods can be listed. Today, both individual and institutional investors have progressively employed machine and deep learning methodologies as instruments for guiding their investment strategies within financial markets. ...
Investing in financial markets requires an adequately planned approach and decision-making process for both individual and institutional investors. The volatility of financial markets is influenced by intricate and constantly evolving factors, prompting investors, analysts, and financial experts to employ progressively sophisticated and data-centric methodologies to precisely forecast future price swings. Deep learning models for stock price prediction demonstrate the ability to comprehend intricate connections by amalgamating extensive datasets. The objective of this essay is to employ various machine learning models using daily data from the BIST100 index, a prominent financial indicator in Turkey. The models under question encompass Support Vector Regression (SVR), K-Nearest Neighbors (KNN), Random Forest (RF), XGBoost and Stacked Generalization. The models' prediction skills were evaluated using RMSE, MSE, MAE, and R2 performance indicators. Based on the observed results, the Stacked Generalization model demonstrated greater performance in making predictions for the analyzed dataset. These findings offer valuable insights that should be considered when selecting models for similar analyses in the future.
... Contemporary market trend prediction methodologies can be categorized into two primary approaches. Machine learning methods, characterized by pattern recognition from historical data, demonstrate performance limitations and lack interpretability, failing to account for dynamic distributions and complex interactions in real-world markets [11,12,13,14,15]. Simulation methods, featuring agent-based action generation through observational learning and predefined models, offer enhanced interpretability in financial market trend projection [16,17,18]. ...
After decades of evolution, the financial system has increasingly deviated from an idealized framework based on theorems. It necessitates accurate projections of complex market dynamics and human behavioral patterns. With the development of data science and machine intelligence, researchers are trying to digitalize and automate market prediction. However, existing methodologies struggle to represent the diversity of individuals and are regardless of the domino effects of interactions on market dynamics, leading to the poor performance facing abnormal market conditions where non-quantitative information dominates the market. To alleviate these disadvantages requires the introduction of knowledge about how non-quantitative information, like news and policy, affects market dynamics. This study investigates overcoming these challenges through rehearsing potential market trends based on the financial large language model agents whose behaviors are aligned with their cognition and analyses in markets. We propose a hierarchical knowledge architecture for financial large language model agents, integrating fine-tuned language models and specialized generators optimized for trading scenarios. For financial market, we develop an advanced interactive behavioral simulation system that enables users to configure agents and automate market simulations. In this work, we take commodity futures as an example to research the effectiveness of our methodologies. Our real-world case simulation succeeds in rehearsing abnormal market dynamics under geopolitical events and reaches an average accuracy of 3.4% across various points in time after the event on predicting futures price. Experimental results demonstrate our method effectively leverages diverse information to simulate behaviors and their impact on market dynamics through systematic interaction.
... S&P 500 using machine learning methods like LSSVM and PSO for forecasting [46]. Neural network models and the utilization of genetic algorithms were initially documented in [47]. The forecasting process involved the integration of artificial neural networks with genetic algorithms, as proposed by K.-j. ...
Accurately predicting socio-economic trends, including stock market behavior, has become increasingly vital for investors, policymakers, and researchers in today's economic growth. This task is particularly challenging in South Asian nations due to the region's economic instability and the unpredictable nature of financial information. This paper aims to predict stock values in five prominent South Asian stock exchanges, namely Karachi (KSE), Nifty50 (NSE), Colombo (CSE), Dhaka (DSE), and Afghanistan, using machine learning methods and daily data from 2018 to 2023. To improve forecasting accuracy, this research used a greedy approach to optimize the window size of a Simple Moving Average (SMA) and normalized the data to train three deep learning models: Convolutional Neural Networks (CNN), Gated Recurrent Units (GRU), and Long Short-Term Memory (LSTM). The models were evaluated using performance metrics such as Mean Squared Error (MSE), Mean Absolute Error (MAE), and R2 score. Our results demonstrate that GRU outperforms LSTM and CNN in all markets, with reduced MSE and elevated R² values. However, CNN exhibits the most volatility in unstable markets, such as Afghanistan and Sri Lanka. LSTM provides more dynamic forecasting patterns but is prone to overestimating abrupt fluctuations in stock values. In summary, our research provides a comprehensive evaluation of machine learning models for stock price prediction and identifies GRU as the most reliable model.
... Taskaya-Temizel and Casey (2005) also evaluated hybrid ANN techniques for time series forecasting, using diverse random and gradient search algorithms to optimize ANN models based on specific data characteristics. Kim and Han (2000) introduced a method combining Genetic Algorithms with RNA models to forecast stock market indices. ...
This study presents a comprehensive literature review on integrating exogenous variables in time series forecasting, with a particular focus on financial markets. Both traditional statistical methods, such as ARIMA and SARIMA, and machine learning techniques, like neural networks and support vector regression, are systematically examined, emphasizing their impact on prediction accuracy and the practical challenges involved. Significant research gaps are identified through an in-depth analysis of the existing literature, especially regarding the application of hybrid models and advanced computational techniques. Unlike previous reviews, this paper highlights the potential of combining traditional and machine learning approaches to handle multidimensional data from various external sources better. Addressing these gaps could enhance the robustness and precision of financial forecasting models, offering valuable insights to academic researchers and industry practitioners navigating volatile market environments. The paper consolidates current knowledge and underscores promising computational opportunities that could transform time series forecasting.
... Having machines learn huge sets of data such as historical stock prices, trading volumes, accounting performances, fundamental features of the stocks, and even the weather, and produce the future values of stocks or index is one big branch of stock market forecasting methods. It utilizes many learning, regression, classification, neural networks algorithms such as support vector machine, random forest, logistic regression, naive Bayes, and reccurent neural networks, and tries to make accurate predictions by adjusting itself according to the market changes [6,7,8,9]. Another popular method is to use natural language processing techniques that let machines extract and understand information written and spoken in human languages, and try to capture stock market sentiments for making investment decisions based on the mood or the sentiments of the stock market [10,11]. Traditional finance and modern financial engineering also attempt to forecast the stock market using the fundamental and technical analysis. ...
A stock market is considered as one of the highly complex systems, which consists of many components whose prices move up and down without having a clear pattern. The complex nature of a stock market challenges us on making a reliable prediction of its future movements. In this paper, we aim at building a new method to forecast the future movements of Standard & Poor's 500 Index (S&P 500) by constructing time-series complex networks of S&P 500 underlying companies by connecting them with links whose weights are given by the mutual information of 60-minute price movements of the pairs of the companies with the consecutive 5,340 minutes price records. We showed that the changes in the strength distributions of the networks provide an important information on the network's future movements. We built several metrics using the strength distributions and network measurements such as centrality, and we combined the best two predictors by performing a linear combination. We found that the combined predictor and the changes in S&P 500 show a quadratic relationship, and it allows us to predict the amplitude of the one step future change in S&P 500. The result showed significant fluctuations in S&P 500 Index when the combined predictor was high. In terms of making the actual index predictions, we built ARIMA models. We found that adding the network measurements into the ARIMA models improves the model accuracy. These findings are useful for financial market policy makers as an indicator based on which they can interfere with the markets before the markets make a drastic change, and for quantitative investors to improve their forecasting models.
... However, the methods commonly employed for a long time involve models that seek to establish a relationship between historical behavior and future price movements. Forecasts are made about future stock prices using samples from historical market data [2]. In today's stock market forecasting, soft computing artificial intelligence (AI) models like machine learning and hybrid models are extensively utilized. ...
The most important goal for people trading in stock markets is to earn a profit. This profit can be obtained from price increases or decreases in a two-sided market, where stockholders gain profit from the difference in purchase and sale prices. This research aims to provide an integrated algorithmic trading system to improve the Sharpe ratio performance index for selecting stocks and maximizing returns during market ups and downs. The proposed strategy utilizes word2vector, autoencoder, and long short-term memory-artificial neural network (LSTM–ANN) methods. We assume that as the markets interact with the news, an internal connection is created between economic and political news and forecasting the market's price trend. The trading system is developed based on price and news to extract features that predict negative or positive market reactions. The tests show that the proposed Price and Price&News trading systems are superior to the Buy&Hold strategy in the S&P500 market. This research studies price data and portfolio news consisting of 15 stocks. The Price&News strategy yielded a 23% higher return than the Buy&Hold strategy, accompanied by a 2.6 improvement in the Sharpe ratio. Additionally, it outperformed the Price algorithm with an 8% higher return and a 0.82 improvement in the Sharpe ratio.
... Ding et al. (2015) employed five predictive models based on convolutional neural networks (CNN). Some studies have combined multiple methodologies; Kim and Han (2000) proposed an approach that applies genetic algorithms (GA) to artificial neural networks (ANN) to reduce complex dimensions and noise. ...
... There is some concern in the literature about the number of observations needed to obtain good predictions, with some experimentation in this regard through performance measurement and variation of error metrics [31,32], finding significant differences between the observations considered necessary. ...
Deep learning techniques have significantly advanced time series prediction by effectively modeling temporal dependencies, particularly for datasets with numerous observations. Although larger datasets are generally associated with improved accuracy, the results of this study demonstrate that this assumption does not always hold. By progressively increasing the amount of training data in a controlled experimental setup, the best predictive metrics were achieved in intermediate iterations, with variations of up to 66% in RMSE and 44% in MAPE across different models and datasets. The findings challenge the notion that more data necessarily leads to better generalization, showing that additional observations can sometimes result in diminishing returns or even degradation of predictive metrics. These results emphasize the importance of strategically balancing dataset size and model optimization to achieve robust and efficient performance. Such insights offer valuable guidance for time series forecasting, especially in contexts where computational efficiency and predictive accuracy must be optimized.
... then he compared the output of the model of Fuzzy logic with basic back propagation algorithm. He concluded that the neuro-Fuzzy model was able to recognize the general characteristics of the stock market better than the back propagation algorithm [12], utilized of neural network modified to predict the stock price index, the method used aimed to use globally searched future dieselization to reduce the dimensionally of future space, mitigate the limitations gradient descent, and illuminates irrelevant factors. ...
The purpose of this paper is to consider the potential in the projection of Fuzzy logic and Neural networks, also to make some combination between models to address implementation issues in the prediction of index and prices for Amman stock exchange in different models, where the previous researchers have to demonstrate the differences between these measures. We have used in this research Amman stock Exchange index prices data as a sample set to compare the different application models, where predicting the stock market
was very difficult since it depends on nonstationary financial data, in addition to the most of the models are nonlinear systems. These papers draw an existing academic and practitioner in literature review as a combination of these models and compare them, the facilities of the development of conceptual methods and the research proposition are the basis for serving this combination. Hence, the present and recent papers can serve the further researchers into addressing contemporary barriers in the direction of these researchers.
The authors show in this paper the Fuzzy logic and Neural networks, in
addition to time series analysis through these models, utilized of RSI, OS, MACD, and OBV, then using MSE, MAPE, and RMSE. The research implication represents of too much data for the period of study, also this paper is conceptual in its nature, the paper highlights in finding that the implementation challenges, and how these challenges can facilitate the trader decision in the stock market. The results of the analysis show that the ANFIS is the better model to achieve prediction of stock market more than others. When are MAPE and RMSE the best more than simulating the errors in other methods? Also the fuzzy-neural models as the results of table show that more prominent in fuzzy-neural models ,while it appears that in MSE as medium, MAD posses less amount than other models in all table testing fuzzy-neural models, therefore, it becomes superior in stock prediction.
... Compared to machine learning techniques, deep learning which can process non-linear and dynamic financial time series has a better performance [6,7]. The well-known deep learning techniques include Artificial Neural Networks (ANN) [8,9], Recurrent Neural Networks (RNN) [10], Long Short-Term Memory (LSTM) [11], and so on. Yet despite there are certain advantages of those neural networks, they are still unable to forecast the fluctuation of the stock market accurately because there is no regression and traditional neural networks have only shallow architecture. ...
Long-short term memory (LSTM) is a state-of-art and widely used model to forecast financial time series. However, primitive LSTM networks do not perform well due to over-fitting problems of the deep learning model and nonlinear and non-stationary characteristics of financial time series data. In addition, complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) is an outstanding data frequency decomposition technique that can decompose original time series into several intrinsic mode functions and a residue. Thus, this paper proposed a novel hybrid network CEEMDAN-LSTM-BN based on LSTM. Specifically, to avoid over-fitting, the modified LSTM-BN network consists of two LSTM layers, two Batch Normalization (BN) layers following each LSTM layer, and a dropout layer. Each of the intrinsic mode functions and the residue would be processed by CEEMDAN-LSTM-BN and the final prediction results are obtained by reconstructing each predictive series. The advantages of the proposed CEEMDANLSTM-BN networks are verified by comparing them to primitive LSTM, other hybrid models, and some famous machine learning models. Moreover, the robustness of the networks is assessed by numerical experiments on different stock indices datasets.
... The genetic algorithm (GA) is inspired by the principles of Genetics and Natural Selection and is popularly used for finding optimum solutions to various problems. The first step in the GA technique is to propose a set of random populations where each individual is called a chromosome, and each chromosome consists of a fixed length of strings, where a string is called a gene 69,70 . Then, the chromosomes of the first generation (1 st set of population) follow several steps, including selection, cross-over and mutation to create a new population. ...
This research offers a novel methodology for quantifying water needs by assessing weather variables, applying a combination of data preprocessing approaches, and an artificial neural network (ANN) that integrates using a genetic algorithm enabled particle swarm optimisation (PSOGA) algorithm. The PSOGA performance was compared with different hybrid-based metaheuristic algorithms’ behaviour, modified PSO, and PSO as benchmarking techniques. Based on the findings, it is possible to enhance the standard of initial data and select optimal predictions that drive urban water demand through effective data processing. Each model performed adequately in simulating the fundamental dynamics of monthly urban water demand as it relates to meteorological variables, proving that they were all successful. Statistical fitness measures showed that PSOGA-ANN outperformed competing algorithms.
... In recent days, the analysis of the stock market has been done based on historical market prices using certain techniques that involve Neural Networks [20], Case-based reasoning [22], Autoregressive, moving average [18], Support Vector Machines [21], Genetic Algorithm [19], and other strategies for assessing the behavior of the stock market. The issue with an accurate prediction of these techniques is formulating the arbitrary behavior of the market, but there is no explanation for it. ...
... Time series techniques have been extensively explored for modeling and predicting financial markets. Autoregressive models like ARIMA have been commonly used for stock return forecasting Kim & Han (2000). Volatility modeling is also critical in finance, with models like GARCH and its variants applied for risk estimation Bollerslev (1986). ...
In the dynamic world of financial markets, accurate price predictions are essential for informed decision-making. This research proposal outlines a comprehensive study aimed at forecasting stock and currency prices using state-of-the-art Machine Learning (ML) techniques. By delving into the intricacies of models such as Transformers, LSTM, Simple RNN, NHits, and NBeats, we seek to contribute to the realm of financial forecasting , offering valuable insights for investors, financial analysts, and researchers. This article provides an in-depth overview of our methodology, data collection process, model implementations, evaluation metrics, and potential applications of our research findings. The research indicates that NBeats and NHits models exhibit superior performance in financial forecasting tasks, especially with limited data, while Transformers require more data to reach full potential. Our findings offer insights into the strengths of different ML techniques for financial prediction, highlighting specialized models like NBeats and NHits as top performers-thus informing model selection for real-world applications. To enhance readability, all acronyms used in the paper are defined below:
... In the work of [6], authors have created an artificial neural network that employs genetic algorithms to update the weights and biases and observed an increase in accuracy and a decrease in training the network. A fellow hybrid model that uses genetic algorithms in neural networks was used in feature discretization and connection weights rather than using genetic algorithms to optimize the model [7]. Genetic algorithms in general are computationally expensive and are not traditionally used to work with highly volatile data. ...
With the increase in computing power and the popularity of machine learning (ML), it has become the norm to tackle more complex problems using ML. The stock market is known to be a highly volatile environment in which stock prices can fluctuate in an erratic manner. The main goal behind this study is to use a deep learning artificial intelligence model to understand and forecast future stock prices. An analysis was also done to assess the role of social media in the stock market price variation and to what extent, it impacts stock prices. The favored approach was to use a Recurrent neural network (RNN) composed of a Long Short-Term Memory (LSTM) model to predict the prices as it is the most suitable to work with time-series data. A successful model was deployed which showed a high level of accuracy and produced low values with regards to the loss function.
... In the year 2000 great Scholar K. J. Kim et al [1] projected new hybrid model of Artificial Neural Network and Genetic Algorithm for attribute discretization. Attribute discretization concepts to convert continuous data into discrete data using certain thresholds. ...
This paper aims to present a hybrid model to forecast stock price by analyzing different trends of stock market. As the stock price are time series but they are not static and highly noise due to the fact that stock market is not stable as it depends on various factors. In this paper we have intend a new move towards forecast stock price using ANN model optimized by particle swam optimization (PSO) this model is consisting of an effective algorithm for predicting next day high price of Yahoo stock value and Microsoft stock value. To present this algorithm we have taken real dataset of Yahoo Company and Microsoft Company. This new approach is compared with existing models with real data set and gives more accurate results which give more accuracy result with MAPE of 1%. The stock data are most challenging as because they are dynamic, non-liner and no common parameter. So it's very difficult for the inverters to predict stock price without analyzing trends of stock price. The selection and execution of an appropriate forecasting methodology has always been a vital planning and control problem for most firms and agencies. The organizational and monetary permanence of an organization depends on the accurateness of the prediction as such information determines to make key decisions in the areas of possessions, acquisition, advertising, planning and development of any organization and firm. An artificial neural network is defined as an information processing model inspired by the methods by which mammalian brain progress information represent to it. An artificial neural network is a collection of mathematical models which imitate some of the empirical phenomenon in a genetic nervous system, most prominently adaptive natural learning. One distinctive and significant property of artificial neural network model is the excellent structure of the information processing system. It is made of a number of highly interconnected processing elements that are very similar to neurons and are joined by weighted connections that are very similar. In this paper we have proposed an efficient hybrid model which works on time series data to predict stock price and result is also optimized using particle swam optimization. Finally the result is compared with existing models and our model seems to be working more accurately in certain conditions.
The Saudi stock market holds a significant position in the Middle East’s financial landscape, attracting global investors due to its status as the tenth-largest market worldwide by market value. However, forecasting stock prices in this market presents notable challenges owing to its intricate nature and volatility. This study aims to construct a predictive model capable of forecasting Saudi stock prices for the following day, leveraging diverse datasets and algorithms. Specifically, we apply machine learning algorithms for stock price prediction in the Saudi market, focusing on datasets from four prominent companies: Almarai, Saudi Telecommunication Company (STC), Sabic, and Alrajhi Bank. Spanning from 2018 to 2022, our analysis employs various algorithms including linear regression, random forest, XGB, and AdaBoost. Through this research, we aim to assess the accuracy of these models and determine the factors influencing their performance. Our findings consistently demonstrate that linear regression outperforms other models across all datasets, followed by random forest and XGB. Notably, AdaBoost consistently exhibits the lowest accuracy among the models tested. Further data analysis uncovers varying distributions and skewness across datasets, exerting influence on model performance. While machine learning models exhibit promise in stock price prediction, their efficacy is subject to various factors, necessitating ongoing research to refine accuracy and applicability.
The endeavor of stock trend forecasting is principally focused on predicting the future trajectory of the stock market, utilizing either manual or technical methodologies to optimize profitability. Recent advancements in machine learning technologies have showcased their efficacy in discerning authentic profit signals within the realm of stock trend forecasting, predominantly employing temporal data derived from historical stock price patterns. Nevertheless, the inherently volatile and dynamic characteristics of the stock market render the learning and capture of multi-scale temporal dependencies and stable trading opportunities a formidable challenge. This predicament is primarily attributed to the difficulty in distinguishing real profit signal patterns amidst a plethora of mixed, noisy data. In response to these complexities, we propose a Multi-Scale Temporal Memory Learning and Efficient Debiasing (MTMD) model. This innovative approach encompasses the creation of a learnable embedding coupled with external attention, serving as a memory module through self-similarity. It aims to mitigate noise interference and bolster temporal consistency within the model. The MTMD model adeptly amalgamates comprehensive local data at each timestamp while concurrently focusing on salient historical patterns on a global scale. Furthermore, the incorporation of a graph network, tailored to assimilate global and local information, facilitates the adaptive fusion of heterogeneous multi-scale data. Rigorous ablation studies and experimental evaluations affirm that the MTMD model surpasses contemporary state-of-the-art methodologies by a substantial margin in benchmark datasets.
This work aims to present the development of a chatbot (or virtual assistant) that aims to collaborate with the management of its users' tasks. Taking advantage of the concepts of recurrent neural networks and natural language processing, emphasis was placed on developing the code that dictates all the assistant's reactions based on the classification of the message received. The research presents an approach that has previously been little explored within the context of personal productivity, bringing conversation as an alternative for handling appointments and sending alerts, with the potential to be integrated into messaging applications. The design of the tool was possible due to the use of the Google Colaboratory online platform, the TensorFlow framework and the Python programming language.
The purpose of this study is to design a predictive model for a daily quality assurance (QA) system that remains unaffected by specific patterns in correlated time series data. All data were sampled from the measured output factor at specific times over a 5-year period during the daily QA process for a 6 MV photon beam of the Varian linear accelerator (LINAC) system. Before constructing predictive structures, an autocorrelation function (ACF) analysis was conducted to verify the correlation of the given time series data. This study determined the optimal configuration for the autoregressive integrated moving average (ARIMA) and nonlinear autoregressive (NAR) neural network models for prediction. Additionally, it utilized correlated time series data to evaluate its impact on the predictive capability. We then compared the actual QA values to those predicted by the selected ARIMA and NAR models for the sampled daily output. Our findings suggest that while the ARIMA model offers a quick and relatively easy approach without requiring complex computational methods, the NAR model outperforms ARIMA, especially in the context of correlated time series data, demonstrating its real clinical utility as a prediction model. This result reveals that correlations are frequently observed in daily QA data. We concluded that these correlations can substantially influence the accuracy of machine behavior predicted based on historical observations. Consequently, analyzing specific patterns and correlated data is imperative for designing predictive structures.
This study presents a comprehensive literature review exploring the integration of exogenous variables in economic time series forecasting, focusing on supporting public development policies. It systematically examines various computational parametrization techniques incorporating these external variables, emphasizing their potential to enhance prediction accuracy and address practical challenges in policy-driven contexts. By meticulously analyzing the existing body of research, this review identifies significant gaps and outlines future opportunities for advancing computational techniques and model development. Addressing these gaps could lead to more robust and precise economic forecasting models, offering valuable insights for both academic researchers and policymakers in the formulation and implementation of effective public development strategies. This paper consolidates current knowledge and paves the way for innovative computational approaches that can revolutionize economic forecasting by integrating multidimensional data from diverse external sources.
Artificial Intelligence (AI) has been developing rapidly in recent years, and the application of AI in various fields has gradually emerged. This paper focuses on exploring the cross-fertilization of AI with the Chinese stock market, and the study adopts 7 types of factors, totaling 29 factor indicators, covering multiple types of value, valuation, leverage, financial quality, growth, technology, and risk factors, etc. Through data preprocessing, feature engineering, and other steps on the factors, and then combined with 8 machine learning algorithms to construct corresponding quantitative trading strategies on CSI 300 constituent stocks. By comparing the results of various models, it is found that the trading strategies constructed by machine learning algorithms can obtain significant excess returns in the Chinese market. Except for Gaussian Park Bayes, the other seven model strategy returns beat the benchmark returns, among which XGBoost performs the best, achieving a 20.10% return and the smallest retraction. This paper has a positive impact on advancing the research on the intersection of machine learning and quantitative investment.
Predicting the stock market is one of the significant chores and has a successful prediction of stock rates, and it helps in making correct decisions. The prediction of the stock market is the main challenge due to blaring, chaotic data as well as non-stationary data. In this research, the support vector machine (SVM) is devised for performing an effective stock market prediction. At first, the input time series data is considered and the pre-processing of data is done by employing a standard scalar. Then, the time intrinsic features are extracted and the suitable features are selected in the feature selection stage by eliminating other features using recursive feature elimination. Afterwards, the Long Short-Term Memory (LSTM) based prediction is done, wherein LSTM is trained to employ Aquila circle-inspired optimization (ACIO) that is newly introduced by merging Aquila optimizer (AO) with circle-inspired optimization algorithm (CIOA). On the other hand, delay-based matrix formation is conducted by considering input time series data. After that, convolutional neural network (CNN)-based prediction is performed, where CNN is tuned by the same ACIO. Finally, stock market prediction is executed utilizing SVM by fusing the predicted outputs attained from LSTM-based prediction and CNN-based prediction. Furthermore, the SVM attains better outcomes of minimum mean absolute percentage error; (MAPE) and normalized root-mean-square error (RMSE) with values about 0.378 and 0.294.
The current quality assurance (QA) processes are conducted on an hourly, daily, and monthly basis, and these labor-intensive activities increase the demand for predictive models. The purpose of this study is to design a predictive structure for QA. We utilized the nonlinear autoregressive (NAR) neural network to predict the time series. To evaluate the capability of the predictive model, this study involved comparing the actual QA values to the values predicted by the NAR model for the sampled daily output. In predicting the daily output, our study determined that the optimal configuration for the NAR model included 1 input, 1 hidden layer with 20 neurons, and 1 output layer with a single output neuron. Considering the mean absolute percentage error (MAPE) as the evaluation criterion, the NAR model is 1.385. This indicates that the NAR model has clinical utility as a prediction model. The NAR model predicted within ± 0.2% compared to the actual output. Therefore, this study recommends the implementation of a predictive structural model as the next step in enhancing the QA system.
The stock market is characterized by its complexity, dynamism, and sensitivity to a multitude of factors, making accurate trend analysis a paramount concern for investors and traders. This research investigates the application of machine learning techniques for stock market trend analysis, providing a comprehensive study of historical stock prices, economic indicators, and advanced machine learning algorithms. Ensemble methods, particularly Gradient Boosting, outperformed other models in accuracy, precision, recall, and F1-Score. Technical indicators and lag features play a pivotal role in capturing trends, providing actionable insights. The analysis emphasizes the significance of time horizons in model performance, emphasizing the necessity to align model choices with investment strategies. This research advances stock market analysis, demonstrating the value of machine learning predictions for investors and traders
This study contributes to the literature on AI-based stock market forecasting by highlighting the potential of AI-based stock market forecasting in identifying investment opportunities in uncertain times. Globalization and internet access have facilitated portfolio diversification across regions and this paper examines the application of artificial intelligence (AI) in identifying the best global market indices for investment during a recession. The study evaluates markets such as China, the United States, and Japan, as well as emerging markets such as Brazil and Mexico with the aim of making an immediate investment decision. The results suggest that support vector machine and random forest techniques are effective for stock price forecasting, and the S&P BMV IPC Index has been identified as an attractive option. However, real-time data may influence this result, and NASDAQ may offer a better outlook. The research highlights the value of AI-based forecasting in enabling informed investment decisions, portfolio diversification, and capturing opportunities across geographies during downturns.
When attempting to determine the optimal complexity of a neural network the correct decision is dependent upon obtaining the global solution at each stage of the decision process. Failure to ensure that each optimization being considered is near a global solution may lead to misleading and often conflicting results jeopardizing any decision rule for choosing an optimal complexity. Here, a genetic algorithm is used for global search and, by modifying the objective function, is used to simultaneously select a parsimonious structure. The chosen structure often eliminates all connections to unnecessary variables and thus identifies irrelevant variables. Several models are examined to forecast the five day relative difference of the S&P 500 index. The first series of models uses only past S&P 500 values while the second series of models uses other explanatory variables. Models with the complete architecture are compared to those with the reduced structure. Based on the preliminary model ana...
Neural networks are on the threshold of becoming assimilated into the mainstream of financial decision making. As a form of machine learning, this technology is especially attractive for developing trading systems which do not rely on human judgment. In this article we describe a neural network-based system for trading S&P 500 index futures that has, in testing, outperformed passive investment in the index. Our approach, the incorporation of several neural networks into a single composite Boolean decision rule system, demonstrates the potential to generate expected investment returns superior to any of the individual networks as well as the index.
The recent surge in activity of neural network research in business is not surprising since the underlying functions controlling business data are generally unknown and the neural network offers a tool that can approximate the unknown function to any degree of desired accuracy. The vast majority of these studies rely on a gradient algorithm, typically a variation of backpropagation, to obtain the parameters (weights) of the model. The well-known limitations of gradient search techniques applied to complex nonlinear optimization problems such as artificial neural networks have often resulted in inconsistent and unpredictable performance. Many researchers have attempted to address the problems associated with the training algorithm by imposing constraints on the search space or by restructuring the architecture of the neural network. In this paper we demonstrate that such constraints and restructuring are unnecessary if a sufficiently complex initial architecture and an appropriate global search algorithm is used. We further show that the genetic algorithm cannot only serve as a global search algorithm but by appropriately defining the objective function it can simultaneously achieve a parsimonious architecture. The value of using the genetic algorithm over backpropagation for neural network optimization is illustrated through a Monte Carlo study which compares each algorithm on in-sample, interpolation, and extrapolation data for seven test functions.
The ability of neural networks to closely approximate unknown functions to any degree of desired accuracy has generated considerable demand for neural network research in business. The attractiveness of neural network research stems from researchers' need to approximate models within the business environment without having a priori knowledge about the true underlying function. Gradient techniques, such as backpropagation, are currently the most widely used methods for neural network optimization. Since these techniques search for local solutions, they are subject to local convergence and thus can perform poorly even on simple problems when forecasting out-of-sample. Consequently, a global search algorithm is warranted. In this paper we examine tabu search (TS) as a possible alternative to the problematic backpropagation approach. A Monte Carlo study was conducted to test the appropriateness of TS as a global search technique for optimizing neural networks. Holding the neural network architecture constant, 530 independent runs were conducted for each of seven test functions, including a production function that exhibits both increasing and diminishing marginal returns and the Mackey-Glass chaotic time series. In the resulting comparison, TS derived solutions that were significantly superior to those of backpropagation solutions for in-sample, interpolation, and extrapolation test data for all seven test functions. It was also shown that fewer function evaluations were needed to find these optimal values.
This article shows that the use of a genetic algorithm can provide better results for training a feedforward neural network than the traditional techniques of backpropagation. Using a chaotic time series as an illustration, we directly compare the genetic algorithm and backpropagation for effectiveness, ease-of-use, and efficiency for training neural networks.
Many supervised machine learning algorithms require a discrete feature space. In this paper, we review previous work on continuous feature discretization, identify defining characteristics of the methods, and conduct an empirical evaluation of several methods. We compare binning, an unsupervised discretization method, to entropy-based and purity-based methods, which are supervised algorithms. We found that the performance of the Naive-Bayes algorithm significantly improved when features were discretized using an entropy-based method. In fact, over the 16 tested datasets, the discretized version of Naive-Bayes slightly outperformed C4.5 on average. We also show that in some cases, the performance of the C4.5 induction algorithm significantly improved if features were discretized in advance; in our experiments, the performance never significantly degraded, an interesting phenomenon considering the fact that C4.5 is capable of locally discretizing features. 1 Introduction Many algorithm...
From the Publisher:
With advanced computer technologies and their omnipresent usage, data accumulates in a speed unmatchable by the human's capacity to process data. To meet this growing challenge, the research community of knowledge discovery from databases emerged. The key issue studied by this community is, in layman's terms, to make advantageous use of large stores of data. In order to make raw data useful, it is necessary to represent, process, and extract knowledge for various applications. Feature Selection for Knowledge Discovery and Data Mining offers an overview of the methods developed since the 1970's and provides a general framework in order to examine these methods and categorize them. This book employs simple examples to show the essence of representative feature selection methods and compares them using data sets with combinations of intrinsic properties according to the objective of feature selection. In addition, the book suggests guidelines for how to use different methods under various circumstances and points out new challenges in this exciting area of research. Feature Selection for Knowledge Discovery and Data Mining is intended to be used by researchers in machine learning, data mining, knowledge discovery, and databases as a toolbox of relevant tools that help in solving large real-world problems. This book is also intended to serve as a reference book or secondary text for courses on machine learning, data mining, and databases.
At the present time a large number of AI methods have been developed in the field of pattern classification. In this paper, we will compare the performance of a well-known algorithm in machine learning (C4.5) with a recently proposed algorithm in the fuzzy set community (NEFCLASS). We will compare the algorithms both on the accuracy attained and on the size of the induced rule base. Additionally, we will investigate how the selected algorithms perform after they have been pre-processed by discretization and feature selection.
This paper adopts the idea of discretising continuous attributes (Fayyad and Irani 1993) and applies it to lazy learning algorithms (Aha 1990; Aha, Kibler and Albert 1991). This approach converts continuous attributes into nominal attributes at the outset. We investigate the effects of this approach on the performance of lazy learning algorithms and examine it empirically using both real-world and artificial data to characterise the benefits of discretisation in lazy learning algorithms. Specifically, we have showed that discretisation achieves an effect of noise reduction and increases lazy learning algorithms' tolerance for irrelevant continuous attributes. The proposed approach constrains the representation space in lazy learning algorithms to hyper-rectangular regions that are orthogonal to the attribute axes. Our generally better results obtained using a more restricted representation language indicate that employing a powerful representation language in a learning algorithm is not always the best choice as it can lead to a loss of accuracy.
A bstract
We present a new approach to evaluation of bankruptcy risk of firms based on the rough set theory. The concept of a rough set appeared to be an effective tool for the analysis of information systems representing knowledge gained by experience. The financial information system describes a set of objects (firms) by a set of multi‐valued attributes (financial ratios and qualitative variables), called condition attributes. The firms are classified into groups of risk subject to an expert's opinion, called decision attribute. A natural problem of knowledge analysis consists then in discovering relationships, in terms of decision rules, between description of firms by condition attributes and particular decisions. The rough set approach enables one to discover minimal subsets of condition attributes ensuring an acceptable quality of classification of the firms analysed and to derive decision rules from the financial information system which can be used to support decisions about financing new firms. Using the rough set approach one analyses only facts hidden in data, it does not need any additional information about data and does not correct inconsistencies manifested in data; instead, rules produced are categorized into certain and possible. A real problem of the evaluation of bankruptcy risk by a Greek industrial development bank is studied using the rough set approach.
uction and feature extraction. Bothare sometimes called feature discovery. Assuming the original set consists of A 1 ; A 2 ; :::; A nfeatures, these variants can be defined below.Feature construction is a process that discovers missing information about the relationshipsbetween features and augments the space of features by inferring or creatingadditional features [5, 7, 6]. After feature construction, we may have additional mfeatures A n+1 ; A n+2 ; :::; A n+m . For example, a new...
Feature selection has been the focus of interest for quite some time and much work has been done. With the creation of huge databases and the consequent requirements for good machine learning techniques, new problems arise and novel approaches to feature selection are in demand. This survey is a comprehensive overview of many existing methods from the 1970's to the present. It identifies four steps of a typical feature selection method, and categorizes the different existing methods in terms of generation procedures and evaluation functions, and reveals hitherto unattempted combinations of generation procedures and evaluation functions. Representative methods are chosen from each category for detailed explanation and discussion via example. Benchmark datasets with different characteristics are used for comparative study. The strengths and weaknesses of different methods are explained. Guidelines for applying feature selection methods are given based on data types and domain characteris...
An abstract is not available.
Recent years have seen the growth in popularity of neural networks for business decision support because of their capabilities for modeling, estimating, and classifying. Compared to other AI methods for problem solving such as expert systems, neural network approaches are especially useful for their ability to learn adaptively from observations. However, neural network learning performed by algorithms such as back-propagation (BP) are known to be slow due to the size of the search space involved and also the iterative manner in which the algorithm works. In this paper, we show that the degree of difficulty in neural network learning is inherent in the given set of training examples. We propose a technique for measuring such learning difficulty, and then develop a feature construction methodology that helps transform the training data so that both the learning speed and classification accuracy of neural network algorithms are improved. We show the efficacy of the proposed method for financial risk classification, a domain characterized by frequent data noise, lack of functional structure, and high attribute interactions. Moreover, the empirical studies also provide insights into the structural characteristics of neural networks with respect to the input data used as well as possible mechanisms to improve the learning performance.
This paper describes the techniques used for categorizing variables in Snout an intelligent assistant for exploratory data analysis of survey and similar data sets that is currently under development.
We begin by reviewing existing work on category formation in data mining which has been mainly concerned with enabling decision
tree programs to handle numeric variables. It is argued that there are other important but neglected aspects of category formation,
notably the formation of new categorizations of nominal variables. We report the limited success achieved in categorizing
variables from survey data using either endogenous methods or exogenous methods that maximise the association with only one
dependent variable. We then describe the categorization technique used in Snout: a procedure that selects a partition that both maximises the number of variables associated with the partitioned variable
and maximises the strength of those associations. We report on the success achieved using this procedure in exploring real
survey data.
The use of neural networks in pattern classification is a relatively recent phenomena. In some instances the nonparametric neural network approach has demonstrated significant advantages over more conventional methods. However, certain of the drawbacks of neural networks have led to interest in the augmentation of the neural network approach with such supporting tools as genetic algorithms (e.g. in support of neural network training). In this paper, we take yet a further step. Specifically, we present an approach for the simultaneous design and training of neural networks by means of a tailored genetic algorithm. We then demonstrate its employment on the problem of the classification of firms with regard to future fiscal well-being (i.e. are they likely to fail or survive). The resulting ontogenic neural network exhibits, we believe, some particularly attractive characteristics.
The existence of numeric data and large numbers of records in a database present a challenging task in terms of explicit concepts extraction from the raw data. The paper introduces a method that reduces data vertically and horizontally, keeps the discriminating power of the original data, and paves the way for extracting concepts. The method is based on discretization (vertical reduction) and feature selection (horizontal reduction). The experimental results show that (a) the data can be effectively reduced by the proposed method; (b) the predictive accuracy of a classifier (C4.5) can be improved after data and dimensionality reduction; and (c) the classification rules learned are simpler.
The escalation of Neural Network research in Business has been brought about by the ability of neural networks, as a tool, to closely approximate unknown functions to any degree of desired accuracy. Although, gradient based search techniques such as back-propagation are currently the most widely used optimization techniques for training neural networks, it has been shown that these gradient techniques are severely limited in their ability to find global solutions. Global search techniques have been identified as a potential solution to this problem. In this paper we examine two well known global search techniques, Simulated Annealing and the Genetic Algorithm, and compare their performance. A Monte Carlo study was conducted in order to test the appropriateness of these global search techniques for optimizing neural networks.
This study presents a hybrid AI (artificial intelligence) approach to the implementation of trading strategies in the S&P 500 stock index futures market. The hybrid AI approach integrates the rule-based systems technique and the neural networks technique to accurately predict the direction of daily price changes in S&P 500 stock index futures. By highlighting the advantages and overcoming the limitations of both the neural networks technique and rule-based systems technique, the hybrid approach can facilitate the development of more reliable intelligent systems to model expert thinking and to support the decision-making processes. Our methodology differs from other studies in two respects. First, the rule-based systems approach is applied to provide neural networks with training examples. Second, we employ Reasoning Neural Networks (RN) instead of Back Propagation Networks. Empirical results demonstrate that RN outperforms the other two ANN models (Back Propagation Networks and Perceptron). Based upon this hybrid AI approach, the integrated futures trading system (IFTS) is established and employed to trade the S&P 500 stock index futures contracts. Empirical results also confirm that IFTS outperformed the passive buy-and-hold investment strategy during the 6-year testing period from 1988 to 1993.
A neuro-fuzzy methodology is described which involves connectionist minimization of a fuzzy feature evaluation index with unsupervised training. The concept of a flexible membership function incorporating weighed distance is introduced in the evaluation index to make the modeling of clusters more appropriate. A set of optimal weighing coefficients in terms of networks parameters representing individual feature importance is obtained through connectionist minimization. Besides, the investigation includes the development of another algorithm for ranking of different feature subsets using the aforesaid fuzzy evaluation index without neural networks. Results demonstrating the effectiveness of the algorithms for various real life data are provided.
This paper describes the techniques used for categorizing variables in Snout an intelligent assistant for exploratory data analysis of survey and similar data sets that is currently under development. We begin by reviewing existing work on category formation in data mining which has been mainly concerned with enabling decision tree programs to handle numeric variables. It is argued that there are other important but neglected aspects of category formation, notably the formation of new categorizations of nominal variables. We report the limited success achieved in categorizing variables from survey data using either endogenous methods or exogenous methods that maximise the association with only one dependent variable. We then describe the categorization technique used in Snout: a procedure that selects a partition that both maximises the number of variables associated with the partitioned variable and maximises the strength of those associations. We report on the success achieved using ...
This paper adopts the idea of discretising continuous attributes (Fayyad and Irani 1993) and applies it to lazy learning algorithms (Aha 1990; Aha, Kibler and Albert 1991). This approach converts continuous attributes into nominal attributes at the outset. We investigate the effects of this approach on the performance of lazy learning algorithms and examine it empirically using both real-world and artificial data to characterise the benefits of discretisation in lazy learning algorithms. Specifically, we have showed that discretisation achieves an effect of noise reduction and increases lazy learning algorithms' tolerance for irrelevant continuous attributes.
The proposed approach constrains the representation space in lazy learning algorithms to hyper-rectangular regions that are orthogonal to the attribute axes. Our generally better results obtained using a more restricted representation language indicate that employing a powerful representation language in a learning algorithm is not always the best choice as it can lead to a loss of accuracy.
In this paper we investigate ways to use prior knowledge and neural networks to improve multivariate prediction ability. Daily stock prices are predicted as a complicated real-world problem, taking non-numerical factors such as political and international events are into account. We have studied types of prior knowledge which are difficult to insert into initial network structures or to represent in the form of error measurements. We make use of prior knowledge of stock price predictions and newspaper information on domestic and foreign events. Event-knowledge is extracted from newspaper headlines according to prior knowledge. We choose several economic indicators, also according to prior knowledge, and input them together with event-knowledge into neural networks. The use of event-knowledge and neural networks is shown to be effective experimentally: the prediction error of our approach is smaller than that of multiple regression analysis on the 5% level of significance. © 1997 by John Wiley & Sons, Ltd.
This article addresses the problem of analyzing existing discretizations of continuous attributes with regard to their redundancy and minimality properties. The research was inspired by the increasing number of heuristic algorithms created for generating the discretizations using various methodologies, and apparent lack of any direct techniques for examining the solutions obtained as far as their basic properties, (e.g., the redundancy), are concerned. The proposed method of analysis fills this gap by providing a test for redundancy and enabling for a controlled reduction of the discretization's size within specified limits. Rough set theory techniques are used as the basic tools in this method. Exemplary results of discretization analyses for some known real-life data sets are presented for illustration.
As of this writing, an large number of AI methods have been developed in the fiel of pattern classification. In this paper, we will compare the performance of a well-known algorithm in machine learning (C4.5) with a recently proposed algorithm in the fuzzy set community (NEFCLASS). We will compare the algorithms both on the attained accuracy as on the size of the induced rule base. Additionally, we will investigate how the selected algorithms perform after they have been pre-processed by discretization and feature selection.
David Goldberg's Genetic Algorithms in Search, Optimization and Machine Learning is by far the bestselling introduction to genetic algorithms. Goldberg is one of the preeminent researchers in the field--he has published over 100 research articles on genetic algorithms and is a student of John Holland, the father of genetic algorithms--and his deep understanding of the material shines through. The book contains a complete listing of a simple genetic algorithm in Pascal, which C programmers can easily understand. The book covers all of the important topics in the field, including crossover, mutation, classifier systems, and fitness scaling, giving a novice with a computer science background enough information to implement a genetic algorithm and describe genetic algorithms to a friend.
We propose the discretization of real-valued financial time series into few ordinal values and use sparse Markov chains within the framework of generalized linear models for such categorical time series. The discretization operation causes a large reduction in the complexity of the data. We analyse daily return and volume data and estimate the probability structure of the process of lower extreme, upper extreme and the complementary usual events. Knowing the whole probability law of such ordinalvalued vector processes of extreme events of return and volume allows us to quantify non-linear associations. In particular, we find a new kind of asymmetry in the return - volume relationship. Estimated probabilities are also used to compute the MAP predictor whose power is found to be remarkably high.
The prediction of stock price performance is a difficult and
complex problem. Multivariate analytical techniques using both
quantitative and qualitative variables have repeatedly been used to help
form the basis of investor stock price expectations and, hence,
influence investment decision making. However, the performance of
multivariate analytical techniques is often less than conclusive and
needs to be improved to more accurately forecast stock price
performance. A neural network method has demonstrated its capability of
addressing complex problems. A neural network method may be able to
enhance an investor's forecasting ability. The purpose of this paper is
to examine the capability of a neural network method and compares its
predictive power with that of multiple discriminant analysis methods
The arbitrage pricing theory (APT) offers an alternative to the
traditional asset pricing model in finance. In almost all of the
literature, a statistical methodology called factor analysis is used to
test or estimate the APT model. The major shortcoming of this procedure
is that it identifies neither the number nor the definition of the
factors that influence the assets. A unique solution to this problem is
offered. It uses a simple back-propagation neural network with a
generalized delta rule to learn the interaction of the market factors
and securities return. This technique can be used to investigate the
effect of several variables on one another simultaneously without being
plagued with uncertainty of probability distributions of each variable
Recurrent neural networks were applied to the recognition of stock
patterns, and a method for evaluating the networks was developed. In
stock trading, triangle patterns indicate an important clue to the trend
of future change in stock prices, but the patterns are not clearly
defined by rule-based approaches. From stock-price data for all names of
corporations listed in the first section of the Tokyo Stock Exchange, an
expert called chart reader extracted 16 triangles. These patterns were
divided into two groups, 15 training patterns and one test pattern.
Using stock data from the past three years for 16 names, 16 recognition
experiments in which the groups were cyclically used were carried out.
The experiments revealed that the given test triangle was accurately
recognized in 15 out of 16 experiments and that the number of the
mismatching patterns was 1.06 per name on the average. A method was
developed for evaluating recurrent networks with context transition
performances, particularly temporal transition performances. The method
for the triangle sequences is applicable to reducing mismatching
patterns
A discussion is presented of a buying- and selling-time prediction
system for stocks on the Tokyo Stock Exchange and the analysis of
internal representation. The system is based on modular neural networks.
The authors developed a number of learning algorithms and prediction
methods for the TOPIX (Tokyo Stock Exchange Prices Indexes) prediction
system. The prediction system achieved accurate predictions, and the
simulation on stocks trading showed an excellent profit
The authors have developed an autonomous system that transforms
feature spaces to improve classification techniques. They apply their
method to an eye-detection face recognition system, demonstrating
substantially better classification rates than competing systems
: The performance of many machine learning algorithms can be substantially improved with a proper discretization scheme. In this paper we describe a theoretically rigorous approach to discretization of continuous attribute values, based on a Bayesian clustering framework. The method produces a probabilistic scoring metric for different discretizations, and it can be combined with various types of learning algorithms working on discrete data. The approach is validated by demonstrating empirically the performance improvement of the Naive Bayes classifier when Bayesian discretization is used instead of the standard equal frequency interval discretization. 1 INTRODUCTION Many algorithms developed in the machine learning and uncertain reasoning community focus on learning in nominal feature bases. On the other hand, many real world tasks involve continuous attribute domains. Consequently, in order to be able to use such algorithms, a discretization process is needed. Continuous variable d...
Financial forecasting is an example of a signal processing problem which is challenging due to small sample sizes, high noise, non-stationarity, and non-linearity. Neural networks have been very successful in a number of signal processing applications. We discuss fundamental limitations and inherent difficulties when using neural networks for the processing of high noise, small sample size signals. We introduce a new intelligent signal processing method which addresses the difficulties. The method uses conversion into a symbolic representation with a self-organizing map, and grammatical inference with recurrent neural networks. We apply the method to the prediction of daily foreign exchange rates, addressing difficulties with non-stationarity, overfitting, and unequal a priori class probabilities, and we find significant predictability in comprehensive experiments covering 5 different foreign exchange rates. The method correctly predicts the direction of change for the next day with an error rate of 47.1%. The error rate reduces to around 40% when rejecting examples where the system has low confidence in its prediction. The symbolic representation aids the extraction of symbolic knowledge from the recurrent neural networks in the form of deterministic finite state automata. These automata explain the operation of the system and are often relatively simple. Rules related to well known behavior such as trend following and mean reversal are extracted.
Trading S&P 500 stock index futures using a neural network
- J H Choi
- M K Lee
- M W Rhee
Choi, J.H., Lee, M.K., & Rhee, M.W. (1995). Trading S&P 500 stock index
futures using a neural network. Proceedings of the Third Annual International Conference on Artificial Intelligence Applications on Wall
Street (pp. 63-72). New York.
Neuro-genetic approach for bankruptcy prediction: a comparison to back-propagation algorithms
- K Shin
- T Shin
- I Han
Shin, K., Shin, T., & Han, I. (1998). Neuro-genetic approach for bankruptcy
prediction: a comparison to back-propagation algorithms. Proceedings
of the International Conference of the Korea Society of Management
Information Systems (pp. 585-597). Seoul, South Korea.
Supervised and unsuper-vised discretization of continuous features Neural network futures trading—a feasibility study, Society for Worldwide Interbank Financial Telecom-munications, Adaptive intelligent systems
- J Dougherty
- R Kohavi
- M Sahami
Dougherty, J., Kohavi, R., & Sahami, M. (1995). Supervised and unsuper-vised discretization of continuous features. Proceedings of the 12th International Conference on Machine Learning (pp. 194–202). San Francisco, CA. Duke, L. S., & Long, J. A. (1993). Neural network futures trading—a feasibility study, Society for Worldwide Interbank Financial Telecom-munications, Adaptive intelligent systems. (pp. 121–132). Amsterdam: Elsevier.
Hybrid neural, genetic, and fuzzy systems
- Wong
Wong, F., & Tan, C. (1994). Hybrid neural, genetic, and fuzzy
systems. In G. J. Deboeck, Trading on the edge (pp. 243-261).
New York: Wiley.
Feature transformation and subset selec-tion. IEEE Intelligent Systems and Their Applications
- H Liu
- H Motoda
Liu, H., & Motoda, H. (1998b). Feature transformation and subset selec-tion. IEEE Intelligent Systems and Their Applications, 13 (2), 26–28.
Investor's guide to technical analysis: predicting price action in the markets
- E Gifford
Gifford, E. (1995). Investor's guide to technical analysis: predicting price
action in the markets, London: Pitman.
Technical indicators and analysis methods
- J Chang
- Y Jung
- K Yeon
- J Jun
- D Shin
- H Kim
Chang, J., Jung, Y., Yeon, K., Jun, J., Shin, D., & Kim, H. (1996). Technical indicators and analysis methods, Seoul: Jinritamgu Publishing.
Modeling structured nonlinear knowledge to predict stock market returns
- Hiemstra
Hiemstra, Y. (1995). Modeling structured nonlinear knowledge to predict
stock market returns. In R. R. Trippi, Chaos and nonlinear dynamics in
the financial markets: theory, evidence and applications (pp. 163-175).
Chicago, IL: Irwin.