Improving the performance of intelligent stock trading systems by using a high level representation for the inputs

Abstract

Intelligent stock trading systems use soft computing techniques for forecasting the trend of the stock price. But the so-called noise in the market usually results in overtrading and loss of profit. In order to reduce the effect of noise on the trading decisions, high level representations can be used for the output of the trading systems. But the technical indicators which act as the inputs of the trading system, suffer from these short term irregularities as well. This paper suggests a high level representation for the technical indicators to match the level of information in the outputs. Digital low pass filters are carefully designed to remove the transient fluctuations of the technical indicators without losing too much information. Several experiments on different stocks in Tehran Stock Exchange shows a major improvement in the performance of the intelligent stock trading systems.