Article: Application of a Modified Generalized Regression Neural Networks Algorithm in Economics and FinanceEleftherios Giovanis[show abstract] [hide abstract]
ABSTRACT: In this paper we propose an alternative and modified Generalized Regression Neural Networks Autoregressive model (GRNN-AR) in S&P 500 and FTSE 100 index returns, as also in Gross domestic product growth rate of Italy, USA and UK. We compare the forecasts with Generalized Autoregressive conditional Heteroskedasticity (GARCH) and Autoregressive Integrated Moving Average (ARIMA) models. The results indicate that GRNN outperform significant the conventional econometric models and can be an efficient alternative tool for forecasting. The MATLAB algorithm we propose is provided in appendix for further applications, suggestions, modifications and improvements.Capital Markets: Asset Pricing & Valuation eJournal. 04/2011;
Article: GARCH - Monte-Carlo Simulation Models with Wavelets Decomposition Algorithm for Stock ReturnsEleftherios Giovanis[show abstract] [hide abstract]
ABSTRACT: In this paper we examine four different approaches in trading rules for stock returns. More specifically we examine the popular procedures in technical analysis, which are the moving average and the Moving Average Convergence-Divergence (MACD) oscillator. The third approach is the simple random walk autoregressive model and the fourth model we propose is a Generalized Autoregressive Conditional Heteroskedasticity (GARCH) regression with wavelets decomposition and Monte- Carlo simulations algorithm developed in MATLAB. We examine five major stock market index returns for a testing forecasting period of 10 days ahead. We conclude that moving average and MACD might lead to net profits, but not in all cases, therefore are not consistent procedures. Furthermore, moving average 1-30 provides the best results. On the other hand random walk autoregressive model leads in all cases to net losses. Finally, the model we propose not only leads always to net profits, but also to significant higher profits in three stock indices than the respective conventional technical analysis toolsInternational Journal of Computer Information Systems. 01/2011; 2.
Article: Study of Discrete Choice Models and Artificial Intelligence Approaches in the Prediction of Economic CrisesEleftherios GiovanisInternational Journal of Computer Information Systems. 01/2011; 2.
Article: Application of Feed-Forward Neural Networks Autoregressive Models in Gross Domestic Product PredictionEleftherios Giovanis[show abstract] [hide abstract]
ABSTRACT: In this paper we present an autoregressive model with neural networks modeling and standard error backpropagation algorithm training optimization in order to predict the gross domestic product (GDP) growth rate of four countries. Specifically we propose a kind of weighted regression, which can be used for econometric purposes, where the initial inputs are multiplied by the neural networks final optimum weights from input-hidden layer after the training process. The forecasts are compared with those of the ordinary autoregressive model and we conclude that the proposed regression’s forecasting results outperform significant those of autoregressive model in the out-of-sample period. The idea behind this approach is to propose a parametric regression with weighted variables in order to test for the statistical significance and the magnitude of the estimated autoregressive coefficients and simultaneously to estimate the forecasts.ERN: Forecasting & Simulation (Production) (Topic). 09/2010;
Article: Applications of Logit and Fuzzy Regressions for the Prediction of Economic Recessions in US EconomyEleftherios Giovanis[show abstract] [hide abstract]
ABSTRACT: In this paper we examine three binary regressions in order to predict the financial crisis or no crisis periods in USA. The first one is the Logit model and the other two are binary fuzzy regressions with sigmoid and triangular membership functions. We apply the models in period 1926-2005 and we examine the forecasting performance for the in-sample period as also in the out-of sample period, which is defined the period 2006-2009. Next we repeat the estimation process for the period 1926-2009 and we apply and examine the prediction and correctly classification percentage for the period 2010. We find that the forecasting performance of fuzzy triangle regression outperforms Logit model in the in-sample period 1926-2005, where the overall correct classification percentage of Logit regression is 80.74 per cent, while the overall forecasting performance of fuzzy regressions with sigmoid and triangular membership functions is respectively 77.19 and 91.09 respectively. Furthermore the fuzzy regressions outperforms significant the forecasting validity of Logit model, as with the last model we find an overall prediction percentage only of 62.50% in the out-of sample period 2006-2009, while with fuzzy regression we get 100.00 per cent forecast percentage. The empirical results indicate that fuzzy regressions provide a better and more reliable signal on whether or not a financial crisis will take place and are able to capture nonlinearities and imprecision than traditional econometric models do. Furthermore, based on the estimated values for the 2010 for the US economy factor we examine, we predict with all models that the economic recession will be continued through 2010.ERN: Other Macroeconomics: Prices, Business Fluctuations, & Cycles (Topic). 09/2010;