Using relative movement to support ANN-based stock forecasting in Thai stock market

Over the years, Artificial Neural Networks (ANNs) have become a popular and seemingly accurate model to forecast stock prices. This paper proposes data preprocessing using relative movement to improve performance of ANN-based stock forecasting. Both fundamental and technical indicators are chosen as inputs to the system. The evaluation metrics include hit ratio and total return. The k-fold cross validation is utilised on a dataset of stocks in the banking sector in the Stock Exchange of Thailand (SET). The experiments show that the proposed model outperforms a traditional model, a random walk model, and a buy & hold strategy for both hit ratio and total return.