Article

MWDINet: A multilevel wavelet decomposition interaction network for stock price prediction

Authors:
To read the full-text of this research, you can request a copy directly from the authors.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... Stock price prediction has long been a focal point of research and practice in the realms of finance and investment [1], [2], [3]. The ability to forecast future movements in stock prices holds significant implications for investors, financial analysts, and policymakers, enabling them to formulate informed decisions, mitigate risks, and capitalize on emerging opportunities within the dynamic landscape of financial markets [4]. ...
Article
Stock price prediction is a critical task in finance and investment, enabling investors to make informed decisions and capitalize on market opportunities. This paper explores the application of predictive modeling techniques to forecast the stock prices of Roblox Corporation, a prominent player in the gaming industry. Despite the growing interest in predictive analytics, there remains a research gap concerning the application of these techniques to specific companies, particularly within the gaming sector. To address this gap, we employ a comprehensive dataset spanning from March 2021 to June 2023, obtained from Yahoo Finance, to develop predictive models using both time series analysis and machine learning algorithms. Our analysis encompasses exploratory data analysis, model development, and evaluation, culminating in insights into Roblox's stock price dynamics and model performance. The evaluation of our predictive models reveals promising results, with a Mean Squared Error (MSE) of 1.22, Root Mean Squared Error (RMSE) of 1.10, and a high R-squared (R2) score of 0.998. These metrics indicate relatively low prediction errors and a strong explanatory power of the models in capturing the variance in Roblox's closing prices. The findings shed light on the unique challenges and opportunities in predicting stock prices within the gaming industry and contribute to the growing body of knowledge in finance and investment. Through our research endeavors, we aim to empower investors and stakeholders with actionable insights to navigate the complexities of financial markets and make informed decisions with confidence and agility.
... An adaptive predictive generative adversarial network model based on graph neural networks has been proposed to mine the spatio-temporal characteristics of winds, producing wind speed scenarios with better diversity, higher accuracy and more realistic shapes (Liu et al., 2024). Some scholars first decompose the wind speed trajectory into low and high frequency components with discrete wavelet transform (Wen et al., 2024) for more accurate prediction by applying deep learning algorithms, and then two recurrent neural network models are employed to efficiently predict the short-term and long-term dependence of the wind speed data and to obtain the most accurate prediction results (Barjasteh et al., 2024). ...
Article
Full-text available
The dependence of wind power on the natural environment leads to volatility, which can cause hidden dangers to the safe and stable operation of the power grid. In this work, a deep learning-based GoogLeNet-embedded no-pooling dimension fully-connected prediction network is proposed for the short-term prediction issue of wind power generation, and the deep learning-based GoogLeNet-embedded no-pooling dimension fully-connected network is compared with five algorithms including long short-term memory network and NasNet. The dataset was collected in Natal. The six algorithms employed predicted the value of wind power for the coming day. Among all, the deep learning-based GoogLeNet embedded no-pooling dimension fully-connected network achieved the optimal prediction results and evaluation metrics. The percentage reduction of each metric value from the second smallest long short-term memory network for the deep learning-based GoogLeNet-embedded no-pooling dimension fully-connected network is 27.0% for mean absolute error, 27.2% for mean absolute percentage error, 34.8% for mean squared error, 19.9% for root mean square error and 21.6% for symmetric mean absolute percentage error.
... Dechun Wen introduced MWDINet, a stock price prediction framework integrating wavelet decomposition, Hull moving average, and autocorrelation correction modules. Experimental results demonstrate its superiority over existing models, showcasing its potential for accurate stock price forecasting [33]. ...
Article
Full-text available
In the dynamic and uncertain stock market, precise forecasting and decision-making are crucial for profitability. Traditional deep neural networks (DNN) often struggle with capturing long-term dependencies and multi-scale features in complex financial time series data. To address these challenges, we introduce CMR-DQN, an innovative framework that integrates discrete wavelet transform (DWT) for multi-scale data analysis, temporal convolutional network (TCN) for extracting deep temporal features, and a GRU–LSTM–Attention mechanism to enhance the model’s focus and memory. Additionally, CMR-DQN employs the Rainbow DQN reinforcement learning strategy to learn optimal trading strategies in a simulated environment. CMR-DQN significantly improved the total return rate on six selected stocks, with increases ranging from 20.37% to 55.32%. It also demonstrated substantial improvements over the baseline model in terms of Sharpe ratio and maximum drawdown, indicating increased excess returns per unit of total risk and reduced investment risk. These results underscore the efficiency and effectiveness of CMR-DQN in handling multi-scale time series data and optimizing stock market decisions.
Article
Due to China’s thriving economy and culture, the performing arts sector has grown remarkably. To study its development, this study has examined the closing prices of performing arts companies. The GA-BPN model was used to analyze the daily closing prices of Funshine Culture (ticker: 300860) and Sanxiang Impression (ticker: 000863) for the predictions of their future daily closing prices. Next, the study compared the predicted prices with the actual closing prices. By comparing four models, namely GA, 7-4-1, 7-4-4-1, and 7-4-4-4-1, the GA-BPN model has a mean square error (MSE) of 2472.580273 and a root mean square error (RMSE) of 49.72504674, which is the smallest value and the smallest error among the four assessment metrics, it was determined that the GA-BPN model yielded the most accurate prediction results, so it was suitable for forecasting stock closing prices.
Article
Full-text available
Forecasting stock movements is a crucial research endeavor in finance, aiding traders in making informed decisions for enhanced profitability. Utilizing actual stock prices and correlating factors from the Wind platform presents a potent yet intricate forecasting approach. While previous methodologies have explored this avenue, they encounter challenges including limited comprehension of interrelations among stock data elements, diminished accuracy in extensive series, and struggles with anomaly points. This paper introduces an advanced hybrid model for stock price prediction, termed PMANet. PMANet is founded on Multi-scale Timing Feature Attention, amalgamating Multi-scale Timing Feature Convolution and Ant Particle Swarm Optimization. The model elevates the understanding of dependencies and interrelations within stock data sequences through Probabilistic Positional Attention. Furthermore, the Encoder incorporates Multi-scale Timing Feature Convolution, augmenting the model's capacity to discern multi-scale and significant features while adeptly managing lengthy input sequences. Additionally, the model's proficiency in addressing anomaly points in stock sequences is enhanced by substituting the optimizer with Ant Particle Swarm Optimization. To ascertain the model’s efficacy and applicability, we conducted an empirical study using stocks from four pivotal industries in China. The experimental outcomes demonstrate that PMANet is both feasible and versatile in its predictive capability, yielding forecasts closely aligned with actual values, thereby fulfilling application requirements more effectively.
Article
This study proposes a novel hybrid model, called ICE2DE-MDL, integrating secondary decomposition, entropy, machine and deep learning methods to predict a stock closing price. In this context, first of all, the noise contained in the financial time series was eliminated. A denoising method, which utilizes entropy and the two-level ICEEMDAN methodology, is suggested to achieve this. Subsequently, we applied many deep learning and machine learning methods, including long-short term memory (LSTM), LSTM-BN, gated recurrent unit (GRU), and SVR, to the IMFs obtained from the decomposition, classifying them as noiseless. Afterward, the best training method was determined for each IMF. Finally, the proposed model’s forecast was obtained by hierarchically combining the prediction results of each IMF. The ICE2DE-MDL model was applied to eight stock market indices and three stock data sets, and the next day’s closing price of these stock items was predicted. The results indicate that RMSE values ranged from 0.031 to 0.244, MAE values ranged from 0.026 to 0.144, MAPE values ranged from 0.128 to 0.594, and R-squared values ranged from 0.905 to 0.998 for stock indices and stock forecasts. Furthermore, comparisons were made with various hybrid models proposed within the scope of stock forecasting to evaluate the performance of the ICE2DE-MDL model. Upon comparison, The ICE2DE-MDL model demonstrated superior performance relative to existing models in the literature for both forecasting stock market indices and individual stocks. Additionally, to our knowledge, this study is the first to effectively eliminate noise in stock item data using the concepts of entropy and ICEEMDAN. It is also the second study to apply ICEEMDAN to a financial time series prediction problem.
Article
Full-text available
This article investigates how uncertainty impacts the effect of monetary policy surprises on stock returns. Using high-frequency US data, we demonstrate that stock markets respond more aggressively to monetary policy surprises during periods of high uncertainty. We also show that uncertainty asymmetrically influences the transmission of positive and negative monetary policy surprises to stock market prices. The amplifying effect of uncertainty is found to be stronger for expansionary shocks than for contractionary shocks. Our robustness analysis confirms that financial uncertainty has a significant role in shaping the influence of monetary policy on the stock market.
Article
Full-text available
We use text-mining techniques to measure the accessibility and quality of information within the texts of interest rate announcements published by the Bank of Israel over the past decade. We find that comprehension of interest rate announcements published by the Bank of Israel requires fewer years of education than interest rate announcements published by the Federal Reserve and the European Central Bank. In addition, we show that the sentiment within these announcements is aligned with economic fluctuations. We also find that textual uncertainty is correlated with the volatility of the domestic financial market.
Article
Full-text available
Each person’s characteristics may influence that person’s behaviors and outcomes. This study builds and uses a new database to estimate experts’ performance and boldness based on their experience and characteristics. Our study classifies experts providing inflation forecasts based on their education, experience, gender, and environment. We provide alternative interpretations of factors affecting experts’ inflation forecasting performance, boldness, and pessimism by linking behavioral economics, the economics of education, and forecasting literature. The study finds that an expert with previous experience at a central bank appears to have a lower propensity for predicting deflation.
Article
Full-text available
We present a hierarchical architecture based on recurrent neural networks for predicting disaggregated inflation components of the Consumer Price Index (CPI). While the majority of existing research is focused on predicting headline inflation, many economic and financial institutions are interested in its partial disaggregated components. To this end, we developed the novel Hierarchical Recurrent Neural Network (HRNN) model, which utilizes information from higher levels in the CPI hierarchy to improve predictions at the more volatile lower levels. Based on a large dataset from the US CPI-U index, our evaluations indicate that the HRNN model significantly outperforms a vast array of well-known inflation prediction baselines. Our methodology and results provide additional forecasting measures and possibilities to policy and market makers on sectoral and component-specific price changes.
Article
Full-text available
The stock market plays a vital role in the economic and social organization of many countries. Since stock price time series are highly noisy, nonparametric, volatility, complexity, nonlinearity, dynamics, and chaos, the stock market prediction is an important issue for investors and professional analysts. In the financial field, stock market prediction is not only an important task but also an important research topic. For different problems, researchers have proposed many prediction methods. Many papers provide strong evidence that stock prices can be predicted from past price data. In this paper, we propose a hybrid prediction model for stock price based on long short-term memory (LSTM) and variational mode decomposition (VMD). We use the variational mode decomposition method to decompose the complex time series of stock prices into several relatively flat, regular, and stable subsequences. Then, we use each subsequence to train the long- and short-term memory neural network and predict each subsequence. Finally, we merge the predicted values of several subsequences to form the predicted results of the stock price complex original time series. To verify fully the method, we selected four experimental data for testing. Compared with the prediction results of various prediction methods, the prediction accuracy of our proposed model is higher. Especially in the R2 index, the experimental effect is very good. The proposed method achieves good results of more than 0.991 on each data set. Therefore, our proposed hybrid prediction model is accurate and effective in forecasting stock prices and has practical significance and reference value.
Article
Full-text available
Smoothing images, especially with rich texture, is an important problem in computer vision. Obtaining an ideal result is difficult due to complexity, irregularity, and anisotropicity of the texture. Besides, some properties are shared by the texture and the structure in an image. It is a hard compromise to retain structure and simultaneously remove texture. To create an ideal algorithm for image smoothing, we face three problems. For images with rich textures, the smoothing effect should be enhanced. We should overcome inconsistency of smoothing results in different parts of the image. It is necessary to create a method to evaluate the smoothing effect. We apply texture pre-removal based on global sparse decomposition with a variable smoothing parameter to solve the first two problems. A parametric surface constructed by an improved Bessel method is used to determine the smoothing parameter. Three evaluation measures: edge integrity rate, texture removal rate, and gradient value distribution are proposed to cope with the third problem. We use the alternating direction method of multipliers to complete the whole algorithm and obtain the results. Experiments show that our algorithm is better than existing algorithms both visually and quantitatively. We also demonstrate our method’s ability in other applications such as clip-art compression artifact removal and content-aware image manipulation.
Article
Full-text available
Uncertainty about an economy’s regime can change drastically around a crisis. An imported crisis such as the global financial crisis in the euro area highlights the effect of foreign shocks. Estimating an open-economy nonlinear dynamic stochastic general equilibrium model for the euro area and the United States including Markov-switching volatility shocks, we show that these shocks were significant during the global financial crisis compared with periods of calm. We describe how US shocks from both the real economy and financial markets affected the euro area economy and how bond reallocation occurred between short- and long-term maturities during the global financial crisis. Importantly, the estimated nonlinearities when domestic and foreign financial markets influence the economy, should not be neglected. The nonlinear behavior of market-related variables highlights the importance of higher-order estimation for providing additional interpretations to policymakers.
Article
Full-text available
Recurrent neural networks (RNNs) have been extraordinarily successful for prediction with sequential data. To tackle highly variable and multi-modal real-world data, we introduce Particle Filter Recurrent Neural Networks (PF-RNNs), a new RNN family that explicitly models uncertainty in its internal structure: while an RNN relies on a long, deterministic latent state vector, a PF-RNN maintains a latent state distribution, approximated as a set of particles. For effective learning, we provide a fully differentiable particle filter algorithm that updates the PF-RNN latent state distribution according to the Bayes rule. Experiments demonstrate that the proposed PF-RNNs outperform the corresponding standard gated RNNs on a synthetic robot localization dataset and 10 real-world sequence prediction datasets for text classification, stock price prediction, etc.
Conference Paper
Full-text available
Deep neural networks have achieved promising results in stock trend prediction. However, most of these models have two common drawbacks, including (i) current methods are not sensitive enough to abrupt changes of stock trend, and (ii) forecasting results are not interpretable for humans. To address these two problems, we propose a novel Knowledge-Driven Temporal Convolutional Network (KDTCN) for stock trend prediction and explanation. Firstly, we extract structured events from financial news, and utilize external knowledge from knowledge graph to obtain event embeddings. Then, we combine event embeddings and price values together to forecast stock trend. We evaluate the prediction accuracy to show how knowledge-driven events work on abrupt changes. We also visualize the effect of events and linkage among events based on knowledge graph, to explain why knowledge-driven events are common sources of abrupt changes. Experiments demonstrate that KDTCN can (i) react to abrupt changes much faster and outperform state-of-the-art methods on stock datasets, as well as (ii) facilitate the explanation of prediction particularly with abrupt changes.
Article
Full-text available
With the continuous growth of wind power integration into the electrical grid, accurate wind power forecasting is an important component in management and operation of power systems. Given the challenging nature of wind power forecasting, various methods are presented in the literature to improve wind power forecasting accuracy. Among them, combining different techniques to construct hybrid models has been frequently reported in the literature. Decomposition-based models are a family of hybrid models that firstly decompose the wind speed/power time series into relatively more stationary subseries, and then build forecasting models for each subseries. In this paper, we present a comprehensive review of decomposition-based wind forecasting methods in order to explore their effectiveness. Decomposition-based hybrid forecasting models are classified into different groups based on the decomposition methods, such as, wavelet, empirical mode decomposition, seasonal adjust methods, variational mode decomposition, intrinsic time-scale decomposition, and bernaola galvan algorithm. We discuss decomposition methods in the context of alternative forecasting algorithms, and explore the challenges of each method. Comparative analysis of various decomposition-based models is also provided. We also explore current research activities and challenges, and identify potential directions for future research on this subject.
Conference Paper
Full-text available
Recent years have witnessed the unprecedented rising of time series from almost all kindes of academic and industrial fields. Various types of deep neural network models have been introduced to time series analysis, but the important frequency information is yet lack of effective modeling. In light of this, in this paper we propose a wavelet-based neural network structure called multilevel Wavelet Decomposition Network (mWDN) for building frequency-aware deep learning models for time series analysis. mWDN preserves the advantage of multilevel discrete wavelet decomposition in frequency learning while enables the fine-tuning of all parameters under a deep neural network framework. Based on mWDN, we further propose two deep learning models called Residual Classification Flow (RCF) and multi-frequecy Long Short-Term Memory (mLSTM) for time series classification and forecasting, respectively. The two models take all or partial mWDN decomposed sub-series in different frequencies as input, and resort to the back propagation algorithm to learn all the parameters globally, which enables seamless embedding of wavelet-based frequency analysis into deep learning frameworks. Extensive experiments on 40 UCR datasets and a real-world user volume dataset demonstrate the excellent performance of our time series models based on mWDN. In particular, we propose an importance analysis method to mWDN based models, which successfully identifies those time-series elements and mWDN layers that are crucially important to time series analysis. This indeed indicates the interpretability advantage of mWDN, and can be viewed as an indepth exploration to interpretable deep learning.
Conference Paper
Full-text available
Stock prices are formed based on short and/or long-term commercial and trading activities that reflect different frequencies of trading patterns. However, these patterns are often elusive as they are affected by many uncertain political-economic factors in the real world, such as corporate performances, government policies, and even breaking news circulated across markets. Moreover, time series of stock prices are non-stationary and non-linear, making the prediction of future price trends much challenging. To address them, we propose a novel State Frequency Memory (SFM) recurrent network to capture the multi-frequency trading patterns from past market data to make long and short term predictions over time. Inspired by Discrete Fourier Transform (DFT), the SFM decomposes the hidden states of memory cells into multiple frequency components, each of which models a particular frequency of latent trading pattern underlying the fluctuation of stock price. Then the future stock prices are predicted as a nonlinear mapping of the combination of these components in an Inverse Fourier Transform (IFT) fashion. Modeling multi-frequency trading patterns can enable more accurate predictions for various time ranges: while a short-term prediction usually depends on high frequency trading patterns, a long-term prediction should focus more on the low frequency trading patterns targeting at long-term return. Unfortunately, no existing model explicitly distinguishes between various frequencies of trading patterns to make dynamic predictions in literature. The experiments on the real market data also demonstrate more competitive performance by the SFM as compared with the state-of-the-art methods.
Chapter
Full-text available
In this chapter, neural networks are used to predict the future stock prices and develop a suitable trading system. Wavelet analysis is used to de-noise the time series and the results are compared with the raw time series prediction without wavelet de-noising. Standard and Poor 500 (S&P 500) is used in experiments. We use a gradual data sub-sampling technique, i.e., training the network mostly with recent data, but without neglecting past data. In addition, effects of NASDAQ 100 are studied on prediction of S&P 500. A daily trading strategy is employed to buy/sell according to the predicted prices and to calculate the directional efficiency and the rate of returns for different periods. There are numerous exchange traded funds (ETF’s), which attempt to replicate the performance of S&P 500 by holding the same stocks in the same proportions as the index, and therefore, giving the same percentage returns as S&P 500. Therefore, this study can be used to help invest in any of the various ETFs, which replicates the performance of S&P 500. The experimental results show that neural networks, with appropriate training and input data, can be used to achieve high profits by investing in ETFs based on S&P 500.
Article
Full-text available
The overall test for lack of fit in autoregressive-moving average models proposed by Box & Pierce (1970) is considered. It is shown that a substantially improved approximation results from a simple modification of this test. Some consideration is given to the power of such tests and their robustness when the innovations are nonnormal. Similar modifications in the overall tests used for transfer function-noise models are proposed
Article
Full-text available
In this paper, a novel algorithm is proposed with application in sleep/awake detection and in multiclass sleep stage classification (awake, non rapid eye movement (NREM) sleep and REM sleep). In turn, NREM is further divided into three stages denoted here by S1, S2, and S3. Six electroencephalographic (EEG) and two electro-oculographic (EOG) channels were used in this study. The maximum overlap discrete wavelet transform (MODWT) with the multi-resolution Analysis is applied to extract relevant features from EEG and EOG signals. The extracted feature set is transformed and normalized to reduce the effect of extreme values of features. A set of significant features are selected by mRMR which is a powerful feature selection method. Finally the selected feature set is classified using support vector machines (SVMs). The system achieved 95.0% of average accuracy for sleep/awake detection. As concerns the multiclass case, the average accuracy of sleep stages classification was 93.0%.
Article
Stock prediction with news released on media platform is helpful for investors to make good investment decisions. Recent researches are generally based on single news view, e.g., headline or body, as a predictive indicator and thus information received is insufficient or incomplete which also lacks of study on market information, then bring low performances of models. In this research, we propose a hierarchical attention network based on attentive multi-view news learning (NMNL) to excavate more useful information from news and the stock market for stock prediction. The core of our approach is a news encoder and a market information encoder. In the news encoder, we learn multi-view news information representation from news headlines, bodies and sentiments by regarding them as three independent parts. We find that the combination of headline, body and sentiment outperforms conventional models on single news view. In the market information encoder, we employ the attention mechanism to capture pivotal news information and combine technical indicators to represent representative market information. In addition, a temporal auxiliary based on Bi-directional Long Short-Term Memory (Bi-LSTM) model is used to generate the contextual market information for stock prediction. Extensive experiments demonstrate the superiority of NMNL, which outperforms state-of-the-art stock prediction solutions.
Article
Stock price prediction is an important and challenging research topic, which has wide application prospects. Correct forecasting results can provide valuable guidance to investors and thus reduce the investment risk. To improve the prediction accuracy and obtain better prediction results, a new stock price prediction model called VML is proposed in this paper. First, the VML model slices the stock price series to obtain multiple window series, then uses variational mode decomposition (VMD) to decompose the window series to obtain multiple subseries. Unlike existing decomposition-based methods, VML decomposes the window series to solve the data leakage problem. Next, model-agnostic meta-learning (MAML) algorithm and long short-term memory (LSTM) network are applied to predict the subseries. A method of dividing the decomposed subseries into multiple tasks is proposed for the purpose of utilizing the MAML algorithm to train the initial parameters of the LSTM with good generalization ability. The initial parameters enable LSTM to fine tune dynamically to fit the latest data distribution of stock price data, which mitigates the impact of concept drift on prediction accuracy. Finally, the VML model merges the prediction results of the subseries to obtain the final predicted stock price. Experimental results on stock datasets of the Chinese Stock Market and the American Stock Market demonstrate that the proposed method improves the accuracy of prediction.
Thesis
Time series are everywhere and exist in a wide range of domains. Electrical activities of manufacturing equipment, electrocardiograms, traffic occupancy rates, currency exchange rates, speech signals, and atmospheric measurements can all be seen as examples of time series. Modeling time series across different domains is difficult. In many cases, it requires enormous effort and a significant amount of prior knowledge to generate highly accurate models tailored to a particular time series domain. In response, an increasing body of research focuses on training neural networks on time series, such that the neural networks learn to model the time series. A common assumption in training neural networks on time series is that the errors at different time steps are uncorrelated. However, due to the temporality of the data, errors are actually autocorrelated in many cases, making the assumption inaccurate. In this thesis, we propose to learn the autocorrelation coefficient jointly with the model parameters in order to adjust for autocorrelated errors and thus improve model performances on time series. We first develop our method for time series regression. Then, extensions are made to three other time series tasks: time series forecasting, time series classification, and anomaly detection. Large-scale experiments with various neural network architectures and datasets from the four time series tasks verify the effectiveness of our method. Results show that our method enhances performance across most of these time series modeling tasks.
Chapter
Stock trend prediction is challenging due to complex stock behavior including high volatility. Leveraging an additional source of information, such as social media, can improve predictions, social media data is highly unstructured. To address these challenges, we propose a Multi-Stage TCN-LSTM Hybrid Attentive Network (MSHAN), which utilizes historical stock data with selected technical indicators and weighted social media information to predict daily stock movement. MSHAN uses a multi-stage architecture and hybrid neural networks to capture long-range sequential context dependencies of time-series data and unstructured textual information. We present results using extensive experiments on the actual market public dataset StockNet and demonstrate that MSHAN outperforms other baselines and can successfully predict the directional movement of daily stock prices. The ablation study results show the effectiveness of each component of MSHAN. The market simulation analysis suggests that the application of MSHAN can enhance trading profits.
Article
Forecasting the stock price is a challenging task due to its complex dynamic behaviors, affected by long-term trends, seasonal changes, cyclical changes, and irregular changes. Although many deep learning techniques have been applied to stock price forecasting, few of them have a deep insight into these complex behaviors. In this work, we propose a four-step hybrid model, named ESTA-Net, to adaptively extract these behavior patterns for stock price forecasting. Firstly, the empirical mode decomposition is applied to decompose a closing price sequence into intrinsic mode functions (IMFs). The goal of this step is to extract multiple quasi-stationary features of different time scales from the historical closing price sequence. Secondly, each IMF is modeled and forecasted by a temporal attention long short term memory (TALSTM) network. The TALSTM network is designed to capture the long term dependency of each IMF. Thirdly, the learned deep representations of IMFs are fed into a scale attention network (SANet), which adaptively selects relevant deep representations of multiple time scale features extracted from the historical price sequence. Finally, these learned deep features are fed into a fully connected layer to predict the future closing price. In addition, to make the proposed model learn the movement direction of the closing price, we propose a novel regularization term, i.e. the direction regularization term, to train the proposed model. This regularization term measures the inconsistency between the predicted movement direction and the actual movement direction of the closing price. Experiments show that the proposed model significantly outperforms benchmark models. Particularly, on seven financial market indices, the proposed model with the direction regularization term achieves the highest POCID (32.04% higher than that of CNN) and the lowest MAPE (37.36% lower than that of DA-RNN).
Article
Stock market prediction has been a classical yet challenging problem, with the attention from both economists and computer scientists. With the purpose of building an effective prediction model, both linear and machine learning tools have been explored for the past couple of decades. Lately, deep learning models have been introduced as new frontiers for this topic and the rapid development is too fast to catch up. Hence, our motivation for this survey is to give a latest review of recent works on deep learning models for stock market prediction. We not only category the different data sources, various neural network structures, and common used evaluation metrics, but also the implementation and reproducibility. Our goal is to help the interested researchers to synchronize with the latest progress and also help them to easily reproduce the previous studies as baselines. Based on the summary, we also highlight some future research directions in this topic.
Article
Modelling and forecasting the stock price constitute an important area of financial research for both academics and practitioners. This study seeks to determine whether improvements can be achieved by forecasting the stock index and volatility with the use of a hybrid model and incorporating the financial variables. We extend the literature of stock market forecasting by applying a hybrid model which combines adaptive wavelet transform (AWT), long short-term memory (LSTM) and ARIMAX-GARCH family models to predict stock index and combines AWT, LSTM and heterogeneous autoregressive model of realized volatility (HAR-RV) model to predict stock volatility for two major indexes in the U.S. stock market including the Dow Jones Industrial Average (DJIA) and Nasdaq Composite (IXIC). The results indicate an overall improvement in forecasting of stock index using the AWT-LSTM-ARMAX-FIEGARCH model with student’s t distribution as compared to the benchmark models. The robust test proves that this model has a higher prediction accuracy in prediction of different time horizons (1-, 10-, 15-, 20-, 30-, and 60-days ahead) for both stock indexes. Also, AWT-LSTM improves ability of HAR (3) X-RV model in prediction of the stocks realized volatility for mentioned time horizons.
Article
Nonlinearity and high volatility of financial time series have made it difficult to predict stock price. However, thanks to recent developments in deep learning and methods such as long short-term memory (LSTM) and convolutional neural network (CNN) models, significant improvements have been obtained in the analysis of this type of data. Further, empirical mode decomposition (EMD) and complete ensemble empirical mode decomposition (CEEMD) algorithms decomposing time series to different frequency spectra are among the methods that could be effective in analyzing financial time series. Based on these theoretical frameworks, we propose novel hybrid algorithms, i.e., CEEMD-CNN-LSTM and EMD-CNN-LSTM, which could extract deep features and time sequences, which are finally applied to one-step-ahead prediction. The concept of the suggested algorithm is that when combining these models, some collaboration is established between them that could enhance the analytical power of the model. The practical findings confirm this claim and indicate that CNN alongside LSTM and CEEMD or EMD could enhance the prediction accuracy and outperform other counterparts. Further, the suggested algorithm with CEEMD provides better performance compared to EMD.
Article
A novel adaptive temporal-frequency network (ATFN), which is an end-to-end hybrid model incorporating deep learning networks and frequency patterns, is proposed for mid- and long-term time series forecasting. Within the framework of the ATFN, an augmented sequence to sequence model is used to learn the trend feature of complicated nonstationary time series, a frequency-domain block is used to capture dynamic and complicated periodic patterns of time series data, and a fully connected neural network is used to combine the trend and periodic features for producing a final forecast. An adaptive frequency mechanism consisting of phase adaption, frequency adaption, and amplitude adaption is designed for mapping the frequency spectrum of the current sliding window to that of the forecasting interval. The multilayer neural networks conduct a transformation similar to the inverse discrete Fourier transform for generating a periodic feature forecast. Synthetic data and real-world data with different periodic characteristics are used to evaluate the effectiveness of the proposed model. The experimental results indicate that the ATFN has promising performance and strong adaptability for long-term time-series forecasting.
Article
Deep learning is well-known for extracting high-level abstract features from a large amount of raw data without relying on prior knowledge, which is potentially attractive in forecasting financial time series. Long short-term memory (LSTM) networks are deemed as state-of-the-art techniques in sequence learning, which are less commonly applied to financial time series predictions, yet inherently suitable for this domain. We propose a novel methodology of deep learning prediction, and based on this, construct a deep learning hybrid prediction model for stock markets—CEEMD-PCA-LSTM. In this model, complementary ensemble empirical mode decomposition (CEEMD), as a sequence smoothing and decomposition module, can decompose the fluctuations or trends of different scales of time series step by step, generating a series of intrinsic mode functions (IMFs) with different characteristic scales. Then, with retaining the most of information on raw data, PCA reduces dimension of the decomposed IMFs component, eliminating the redundant information and improving prediction response speed. After that, high-level abstract features are separately fed into LSTM networks to predict closing price of the next trading day for each component. Finally, synthesizing the predicted values of individual components is utilized to obtain a final predicted value. The empirical results of six representative stock indices from three types of markets indicate that our proposed model outperforms benchmark models in terms of predictive accuracy, i.e., lower test error and higher directional symmetry. Leveraging key research findings, we perform trading simulations to validate that the proposed model outperforms benchmark models in both absolute profitability performance and risk-adjusted profitability performance. Furthermore, model robustness test unveils the more stable robustness compared to benchmark models.
Article
Forecasting a financial asset's price is important as one can lower the risk of investment decision- making with accurate forecasts. Recently, the deep neural network is popularly applied in this area of research; however, it is prone to overfitting owing to limited availability of data points for training. We propose a novel data augmentation approach for stock market index forecasting through our ModAugNet framework, which consists of two modules: an overfitting prevention LSTM module and a prediction LSTM module. The performance of the proposed model is evaluated using two different representative stock market data (S&P500 and Korea Composite Stock Price Index 200 (KOSPI200)). The results confirm the excellent forecasting accuracy of the proposed model. ModAugNet-c yields a lower test error than the comparative model (SingleNet) in which an overfitting prevention LSTM module is absent. The test mean squared error (MSE), mean absolute percentage error (MAPE), and mean absolute error (MAE) for S&P500 decreased to 54.1%, 35.5%, and 32.7%, respectively, of the corresponding S&P500 forecasting errors of SingleNet, while the same for KOSPI200 decreased to 48%, 23.9%, and 32.7%, respectively, of the corresponding KOSPI200 forecasting errors of SingleNet. Furthermore, through the analyses of the trained ModAugNet-c, we found that test performance is entirely dependent on the prediction LSTM module. The contribution of this study is its applicability in various instances where it is challenging to artificially augment data, such as medical data analysis and financial time-series modeling.
Article
Fuzzy cognitive maps (FCMs) have been successfully used to model and predict stationary time series. However, it still remains challenging to deal with large-scale non-stationary time series which have trend and vary rapidly with time. In this paper, we propose a time series prediction model based on the hybrid combination of high-order FCMs with the redundant wavelet transform to handle large-scale non-stationary time series. The model is termed as Wavelet-HFCM. The redundant Haar wavelet transform is applied to decompose original non-stationary time series into multivariate time series, then the high-order FCM is used to model and predict multivariate time series. In learning high-order FCM to represent large-scale multivariate time series, a fast high-order high-order FCM learning method is designed on the basis of ridge regression to reduce the learning time. Finally, summing multivariate time series up yields the predicted time series at each time step. Compared with existing classical methods, the experimental results on eight benchmark datasets show the effectiveness of our proposal, indicating that our prediction model can be applied to various prediction tasks.
Article
To improve the prediction capacity of stock price trend, an integrated prediction method is proposed based on Rough Set (RS) and Wavelet Neural Network (WNN). RS is firstly introduced to reduce the feature dimensions of stock price trend. On this basis, RS is used again to determine the structure of WNN, and to obtain the prediction model of stock price trend. Finally, the model is applied to prediction of stock price trend. The simulation results indicate that, through RS attribute reduction, the structure of WNN prediction model can be simplified significantly with the improvement of model performance. The directional symmetry values of prediction, corresponding to SSE Composite Index, CSI 300 Index, All Ordinaries Index, Nikkei 225 Index and Dow Jones Index, are 65.75%, 66.37%, 65.97%, 65.52% and 66.75%, respectively. The prediction results are better than those obtained by other neural networks, SVM, WNN and RS-WNN, which verifies the feasibility and effectiveness of the proposed method of predicting stock price trend.
Article
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.0 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data.
Conference Paper
This paper improves stock market prediction based on genetic algorithms (GA) and wavelet neural networks (WNN) and reports significantly better accuracies compared to existing approaches to stock market prediction, including the hierarchical GA (HGA) WNN. Specifically, we added information such as trading volume as inputs and we used the Morlet wavelet function instead of Morlet-Gaussian wavelet function in our prediction model. We also employed a smaller number of hidden nodes in WNN compared to other research work. The prediction system is tested using Shenzhen Composite Index data.
Article
This paper proposes a Fast Region-based Convolutional Network method (Fast R-CNN) for object detection. Fast R-CNN builds on previous work to efficiently classify object proposals using deep convolutional networks. Compared to previous work, Fast R-CNN employs several innovations to improve training and testing speed while also increasing detection accuracy. Fast R-CNN trains the very deep VGG16 network 9x faster than R-CNN, is 213x faster at test-time, and achieves a higher mAP on PASCAL VOC 2012. Compared to SPPnet, Fast R-CNN trains VGG16 3x faster, tests 10x faster, and is more accurate. Fast R-CNN is implemented in Python and C++ (using Caffe) and is available under the open-source MIT License at https://github.com/rbgirshick/fast-rcnn.
Article
This paper proposes a framework to implement regression-based tests of predictive ability in unstable environments, including, in particular, forecast unbiasedness and efficiency tests, commonly referred to as tests of forecast rationality. Our framework is general: it can be applied to model-based forecasts obtained either with recursive or rolling window estimation schemes, as well as to forecasts that are model free. The proposed tests provide more evidence against forecast rationality than previously found in the Federal Reserve's Greenbook forecasts as well as survey-based private forecasts. It confirms, however, that the Federal Reserve has additional information about current and future states of the economy relative to market participants. Copyright © 2015 John Wiley & Sons, Ltd.
Article
An estimator of the parameters of a nonlinear time series regression is obtained by using an autoregressive assumption to approximate the variance-covariance matrix of the disturbances. Considerations are set forth which suggest that this estimator will have better small sample efficiency than circular estimators. Such is the case for examples considered in a Monte Carlo study.
Article
Multiresolulion representations are very effective for analyzing the information content of images. We study the properties of the operator which approximates a signal at a given resolution. We show that the difference of information between the approximation of a signal at the resolutions 2 j + l and 2 j can be extracted by decomposing this signal on a wavelet orthonormal basis of L2 (Rn). In L2 (R), a wavelet orthonormal basis is a family of functions (√2j Ψ (2 Jx - π))j,n,ez2+ which is built by dilating and translating a unique functiOn Ψ(x). This decomposition defines an orthogonal multiresolulion representation called a wavelet representation. It is computed with a pyramidal algorithm based on convolutions with quadrature mirror lilters. For images, the wavelet representation differentia1es several spatial orientations. We study the application of this representation to data compression in image coding, texture discrimination and fractal analysis.
Article
We point out that autocorrelated error terms require modification of the usual methods of estimation and prediction; and we present evidence showing that the error terms involved in most current formulations of economic relations are highly positively autocorrelated. In doing this we demonstrate that when estimates of autoregressive properties of error terms are based on calculated residuals there is a large bias towards randomness. We demonstrate how much efficiency may be lost by current methods of estimation and prediction; and we give a tentative method of procedure for regaining the lost efficiency.* We wish to express our thanks for the considerable assistance we have received from Richard Stone.
Article
Wavelet analysis is an exciting new method for solving difficult problems in mathematics, physics, and engineering, with modern applications as diverse as wave propagation, data compression, signal processing, image processing, pattern recognition, computer graphics, the detection of aircraft and submarines and other medical image technology. Wavelets allow complex information such as music, speech, images and patterns to be decomposed into elementary forms at different positions and scales and subsequently reconstructed with high precision. Signal transmission is based on transmission of a series of numbers. The series representation of a function is important in all types of signal transmission. The wavelet representation of a function is a new technique. Wavelet transform of a function is the improved version of Fourier transform. Fourier transform is a powerful tool for analyzing the components of a stationary signal. But it is failed for analyzing the non stationary signal where as wavelet transform allows the components of a non-stationary signal to be analyzed. In this paper, our main goal is to find out the advantages of wavelet transform compared to Fourier transform.
Article
This paper rigorously establishes that standard multilayer feedforward networks with as few as one hidden layer using arbitrary squashing functions are capable of approximating any Borel measurable function from one finite dimensional space to another to any desired degree of accuracy, provided sufficiently many hidden units are available. In this sense, multilayer feedforward networks are a class of universal approximators.
Article
The paper considers a number of problems arising from the test of serial correlation based on the d statistic proposed earlier by the authors (Durbin & Watson, 1950, 1951). Methods of computing the exact distribution of d are investigated and the exact distribution is compared with six approximations to it for four sets of published data. It is found that approximations suggested by Theil and Nagar and by Hannan are too inaccurate for practical use but that the beta approximation proposed in the 1950 and 1951 papers and a new approximation, called by us the a + bdU approximation and based, like the beta approximation, on the exact first two moments of d, both perform well. The power of the d test is compared with that of certain exact tests proposed by Theil, Durbin, Koerts and Abrahamse from the standpoint of invariance theory. It is shown that the d test is locally most powerful invariant but that the other tests are not. There are three appendices. The first gives an account of the exact distribution of d. The second derives the mean and variance to a second order of approximation of a modified maximum likelihood statistic closely related to d. The third sets out details of the computations required for the a + bdU approximation.
Article
Since dynamic regression equations are often obtained from rational distributed lag models and include several lagged values of the dependent variable as regressors, high order serial correlation in the disturbances is frequently a more plausible alternative to the assumption of serial independence than the usual first order autoregressive error model. The purpose of this paper is to examine the problem of testing against general autoregressive and moving average error processes. The Lagrange multiplier approach is adopted and it is shown that the test against the nth order autoregressive error model is exactly the same as the test against the nth order moving average alternative. Some comments are made on the treatment of serial correlation.
Forecast comparisons in unstable environments
  • Giacomini
Stock Price Prediction using Attention-based Multi-input LSTM
  • H Li
  • Y Shen
  • Y Zhu
Multi-scale Two-way Deep Neural Network for Stock Trend Prediction
  • G Liu
  • Y Mao
  • Q Sun
  • H Huang
  • W Gao
  • X Li
N-BEATS: Neural basis expansion analysis for interpretable time series forecasting
  • Oreshkin