ANFIS-BASED ADAPTIVE EXPECTATION MODEL FOR FORECASTING STOCK INDEX

Stock market investors often make their short-term investment decisions according to recent stock information, such as the news before the market opened, yesterday's global stock indices' volatility, or the price fluctuations of the last two days. Greater accuracy by forecasting models is in demand because more accurate predictions will bring more profit to investors. Unfortunately, there are four drawbacks to the conventional time series models: (1) most statistical methods rely on some assumptions about the variables; (2) conventional forecasting models do not provide predicting rules for stock price indices; (3) the rules mined from genetic algorithms and artificial neural networks are not easily understandable, and most rule-based forecasting models generate too many predicting rules for a stock price index; and (4) stock market investors usually make short-term decisions based on recent price fluctuations (the last one or two periods), but most time series models use only the last period of stock price in forecasting. This paper addresses these drawbacks by proposing a new model using expectation equation joining to an adaptive-network-based fuzzy inference system (ANFIS) model [1] to forecast the Heng Seng stock index. To evaluate forecasting performance, three different models, (Chen's model [2], Yu's model 13], and the ANFIS model [1]) are used as comparison models. The experimental results indicate that the proposed model is superior to the listing methods in terms of root mean squared error.