Price Prediction of Stock Index Futures Based on SVM

Though accurately forecasting the price of stock index futures is impossible, it is of great significance if the price´s variation trend can be estimated to a certain extent. In this paper, we adopted a Support Vector Machines method to predict the prices of Stock index futures in the next 5 trading days. First, with an information granulation method, the original data of 3 stock index futures were transformed into a series of fuzzy granules. Then the maximum, medium and minimum values of futures´ opening price in each single granule are all extracted. After utilizing the SVM model to regress the values in fuzzy granules, we came up with the variation range of futures´ price in the next few days. These predicted results are consistent with the actual one, which proves the feasibility of our method.