A Comparative Analysis of Computer Based Forecasting Models Used In Stock Exchange Prediction

Abstract

It is difficult to successfully forecast stock market prices to achieve the best result with minimal input. That’s because stock price forecasting is a complex process that depends on both known and unknown factors. This paper evaluates the different computer based forcasting models and presentes a comparative analysis of the use of these models in stock exchange prediction. The paper used data for the period between January 2010 to December 2015, and  used each of The Markov model, Neuro-fuzzy system, Data mining, Neural Network, ARIMA model, Moving average, Genetic algorithm, and Random walk forecasting models to reveal how they have been used in stock market price forecasting.  Also a combination of Neural Network and ARIMA models were used to form a hybrid model and the outcome of the two methods were compared. R statistical program was used to decompose the time series values into trends, seasonal and random components which gave a deeper insight into the behaviour of the stock exchange market. The result obtained from the different forcasting models including the hybridized models showed that the hybridized models gave the most accurate prediction in stock exchange market forcast compared to the single models.