An efficient modelling approach for forecasting financial time series data using support vector regression and windowing operators

Abstract

Forecasting or predicting stock market price and trend is regarded as a challenging task because of its chaotic nature. Stock market is essentially a nonlinear, non-parametric, noisy and deterministically chaotic system because of liquid money, stock adequacy, human behaviour, news related to the stock market, gambling, international money rate and so on. Since it is an emerging sector of the business and also many people are related to this sector, many researchers and experts have their interest to work on this area in order to describe this chaotic system for pattern recognition purpose of the trend. Many researches have been already done and still studies are ongoing for better solutions. Artificial intelligence and machine learning are the major techniques that have been applied in different studies. The aim of this study is also using machine learning algorithm for recognising the pattern of stock market trend in order to predict stock price. For that, support vector regression is used as machine learning technique and different windowing operators are used as input selection technique or data pre-processing steps. Experiments are undertaken on three different stock indexes; Dhaka Stock Exchange DSE, from Bangladesh, S&P 500 stock index and IBM index.