A Machine Learning Approach to Predicting Market Indices: A Case Study of Colombo Stock Exchange, Sri Lanka

Abstract

Generally, stock prices are chaotic and show both linear and nonlinear behaviors. As a result, the ability of forecasting is notoriously problematic, and represents a major challenge with traditional time series mechanisms; most of the traditional approaches are especially weak in forecasting the future in the highly volatile and unbalanced frameworks under global and local financial depressions. This study is an attempt to develop a new hybrid forecasting approach based on back propagation neural network (BPN) to handle random walk data patterns under high volatility. The proposed methodology was successfully implemented to fulfil the daily demands of the All Share Price Index (ASPI) in Colombo Stock Exchange (CSE) Sri Lanka, from April 2009 to March 2017. The Autoregressive Integrated Moving Average (ARIMA) approach is used as a comparison mode.