Development of a stocks buy sell strategy module using heuristics

Abstract

Over the past decade, statistical techniques, expert systems and software learning techniques have been heavily used in the field of financial market prediction using technical analysis. These researches are in support of the assumption that technical analysis is indeed useful, and that equity returns can be predicted from past returns. Nowadays, more complicated forms of data analysis with use of artificial intelligence such as data mining, cased based reasoning, artificial neural networks, pattern matching and genetic algorithms are being explored and used in trading systems for both technical and fundamental analysis. The evaluation of such artificial trading systems are normally based on their returns. Specifically, a simple 1-point buy sell strategy is normally used in the evaluation of the performance. However, the use of predictions of the future as the soul indicator to buy or sell stocks has limitations that performance rendered shows abnormally high deviations in returns for different stocks which is mainly because of the differences in the behavior of the under-lying stock price movements for each and every stock. In this study, we present a new approach that can produce improved performance for artificial equity trading systems. We adopted a heuristics approach based on dynamic algorithms and developed a Buy Sell Strategic Module that uses predictions of the future to generate buy and sell signals for the artificial equity trading system. We have shown that by using our Buy Sell Indicator, incrementing the accuracy of the prediction renders increasingly augmenting returns with acceptable deviations for different stocks used in the simulation. We have also shown that using a strategy module in artificial trading systems, even with less accurate predictions of the future, reasonable returns can be earned. Finally, we have classified our empirical test results (expected returns) by stocks and by prediction errors (accuracy). This can provide a milest...