Predicting Market Performance with Hybrid Model

In this research, a stock market prediction model was proposed to predict performance of Karachi Stock Exchange (KSE), now merged into Pakistan Stock Exchange (PSX). The model was comprised of four sub-models that were all based on different machine learning technique. Each sub-model used 6 input attributes including fuel price, commodity, foreign exchange, interest rate, general public sentiment and related NEWS. The historical data of the market was also used for predicting the market performance using statistical techniques like Auto-Regressive Integrated Moving Average (ARIMA) and Simple Moving Average (SMA). Support Vector Machine, Radial Basis Function (RBF), Artificial Neural Network’s two variants including Single Layer Perceptron and Multi-layer Perceptron were used to design four different sub-models. The results predicted by all the sub-models were merged in the Hybrid Model. The Hybrid model predicts the market performance on basis of output of each four prediction technique following the majority rule. There were two variants of Hybrid model proposed. Variant I gave about 72.8% accuracy while Variant II gave 95.7% accuracy on training data set. The Hybrid model could not predict better than the results achieved by the MLP based sub-model alone, on test data set. The results suggested that behavior of market can be predicted by using more complex model implementing different machine learning techniques.