Studies On Stock Indices Forecast Based On Text Mining, RBF And AFSA

It is a focus of study on how to accurately forecast stock indices. Because of the influence of a number of factors on it, especially for the effect on some factors that can not be quantized, the forecast of stock index is difficult. Aimed at this problem, studies on theories and methods have been made as follows in this paper,:Firstly, the ANN methods have been used in this paper. The closing indices of Shanghai composite index have been put into two ANN forecast methods respectively based on BP principle and RBF principle. It is found that RBF method forecasts more accurate than BP dose, but neither of them forcast satisfactory. It is also found that the ANN methods have some efficiency like needing long simulation time and sinking into local optima. Sencondly, in order to solve these problems, three intelligent algorithms have been used for optimization of ANN. They are GA, PSO and AFSA. The three optimized ANN is used to forecast Shanghai composite index, and the results show that the AFSA algorithm has done the best. Thirdly, there are many faoctors that have influence on the stock indices, some of them can be quantized, yet others can not. Then nine math indices that mostly influence Shanghai composite index have been inserted in forecast model respectively, and the math indices that perform well are selected and put together to find a optimal combination by data mining. Finally, some factors, which could not be measured by data, such as factors of macro-economy and mood, have been firstly classified and then inserted in a software called weka, which is specially designed for doing REPTree. The results of the procedure are IF-Then rules of a decision tree, in which adjust rate of the forecast results can be obtained. This process called text mining.