Financial Forecasting Using Evolutionary Computational Techniques

Abstract

Financial forecasting or specially stock market prediction is one of the hottest field of research lately due to its commercial applications owing to high stakes and the
kinds of attractive benefits that it has to offer. In this project we have analyzed various evolutionary computation algorithms for forecasting of financial data. The financial data has been taken from a large database and has been based on the stock prices in leading stock exchanges .We have based our models on data taken from Bombay Stock Exchange (BSE), S&P500 (Standard and Poor’s) and Dow Jones
Industrial Average (DJIA). We have designed three models and compared those using historical data from the three stock exchanges. The models used were based on:

1. Radial Basis Function parameters updated by Particle swarm optimization.
2. Radial Basis Function parameters updated by Least Mean Square Algorithm.
3. FLANN parameters updated by Particle Swarm optimization.

The raw input for the experiment is the historical daily open, close, high, low and volume of the concerned index. However the actual input to the model was the parameters derived from these data. The results of the experiment have been depicted with the aid of suitable curves where a comparative analysis of the various models is done on the basis on various parameters including error convergence and the Mean Average Percentage Error (MAPE).

Key Words: Radial Basis Functions, FLANN, PSO, LMS