Behavioristic analysis and comparative evaluation of intelligent methodologies for short-term stock price forecasting

Stock price forecasting has been in the center of interest of the stock market analysts and of the research community during the last four decades. A number of important studies have been conducted dealing primarily with the forecasting effectiveness of various models and methods; however, short-term stock price forecasting has not been sufficiently analyzed, thus creating a need for comparative studies between discriminant standard and advanced techniques. The objectives and the scope of this paper emanates from the above remarks. An integrated computational forecasting system is developed, encompassing representative techniques (multiple regression, exponential smoothing, neural networks and Adaptive Network based Fuzzy Inference System – ANFIS), and a practically exhaustive comparative study of the performance and behavior of these techniques as well as the role of their core parameters in short-term stock price forecasting, is conducted. Interesting conclusions about the forecasting abilities of the tested methodologies are drawn, while the impact of the models' parameters on the methods' behavior and the feasibility for accurate stock price predictions are evaluated.