An artificial neural networks approach for short-term modeling of stock price index.

摘要

The main objective of this thesis is to contribute to the development of Intelligent Systems Methods for modeling several systems that are highly non-linear and uncertain. Specifically, this study presents a non-conventional approach to predict a stock price index using Artificial Neural Networks (ANN), in particular, for the Toronto Stock Market price index (S&P/TSX).;Several economic values for a period of approximately 13 years were used as inputs for the system. Initially, data was trained with one epoch of Adaptive Neuro-Fuzzy Inference System to choose the most suitable one(s). The US/CAD exchange rate, world oil price, gold price, and trading day of the week were used as inputs. Therefore, several parameters for building an ANN paradigm were defined in order to choose the best topology. Once the ANN model was built, it was used to predict next day, next week, two week, and one month values of the S&P/TSX. Another model also included interest rates. A Comparison was made between the two models. Finally, the results were evaluated on five different metrics, including Entropy. (Abstract shortened by UMI.).