Data-Driven Constrained Evolutionary Scheme for Predicting Price of Individual Stock in Dynamic Market Environment

Abstract

Predicting stock price is a challenging problem as the market involve multi-agent activities with constantly changing environment. We propose a method of constrained evolutionary (CE) scheme that based on Genetic Algorithm (GA) and Artificial Neural Network (ANN) for stock price prediction. Stock market continuously subject to influences from government policy, investor activity, cooperation activity and many other hidden factors. Due to dynamic and non-linear nature of the market, individual stock price movement are usually hard to predict. Investment strategies used by regular investor usually require constant modification, remain secrecy and sometimes abandoned. One reason for such behavior is due to dynamic structure of the efficient market, where all revealed information will reflect upon the stock price, leads to dynamic behavior of the market and unprofitability of the static strategies. The CE scheme contains mechanisms which are temporal and environmental sensitive that triggers evolutionary changes of the model to create a dynamic response towards external factors.