Adaptive Selection Of KLSE Stocks Using Neural Networks

Stock is becoming a significant investment tools that contributes towards Malaysia economic growth. Thus it is vital to increase investor's confidence in the Malaysia stock market. In this era of Information Age, artificial intelligence is applied to develop sound investment analysis tools in selecting winning Malaysia stocks. Hence in this study, neural network technology is deployed to build an adaptive neural net trading system, specifically adopting the multilayer feedforward network with backpropagation learning algorithm. A 22-18-2-network architecture of a prediction model is derived from the
developed network simulator to predict the following quarter stock price change, of twenty publicly traded Malaysian companies. A promising classification competency of 80 percent correctness is recorded after the network is iteratively trained for 6000 epochs. This study also indicates that the neural network generated forecasting model is capable of outperforming the statistical model, as recorded by 80 percent neural network accuracy versus 77.3 percent binary logistic regression accuracy. The findings conclude that the neural forecasting ability could be further enhanced. Future research could
incorporate technical analyst approach for a comprehensive stock valuation and also integrates with fuzzy technology to handle imprecise data.