Financial forecasting through data mining using integrated methods

Abstract

To an increasing extent over the past decade, software learning methods including neural networks and case based reasoning have been used for prediction in financial markets and other areas. A systematic approach to knowledge discovery for business cycle forecasting or stock price index must be able to accommodate disparate types of information. In the past, the prediction of financial variables has tended to focus on isolated techniques. Examples lie in models such as regression techniques and Box-Jenkins methods as well as autoregressive conditional heteroscedastic (ARCH) model. Experience with artificial intelligence applications since the early 1980s points toward a multistrategy approach to discovery and prediction. In particular, statistical methods such as factor analysis may be used for exploratory analysis to determine the most salient characteristics behind financial market behavior. The results of such analysis may be used as input into a learning system using implicit knowledge representation (neural network) and case based reasoning. These ideas are presented in the context of a predictive system for a leading indicator of Korea in case study. The results demonstrate the utility of integrating statistical and learning methods for financial forecasting. This study also examines the use of the probabilistic neural network for forecasting a stock price index. By far the most popular type of neural network has been backpropagation. However the advantages of other learning techniques such as the swift response of the probabilistic neural network (PNN) suggests the desirability of adapting other models to the predictive function. A novel architecture lies in an array of PNN models to provide graded forecast of multiple discrete values. This thesis presents a comparative evaluation of various learning models. The concepts are investigated against the backdrop of practical applications involving the prediction of a business cycle indicator and a stock m...