Application Of Neural Networks Based On Monte-Carlo-Adaptation Rule In Index Futures Price Prediction

Hushen300index futures has been on the market for four years, and play a important role in risk distribution of China’s financial market:institutional investors use it to avoid systemic risk and speculators use it to get a higher risk-reward. But the stock market swings and high leverage trait put investors into a big risk, finding a way to forecast the futures price is a center problem for every trader in the futures market. As the technology of Artificial Intelligence, artificial neural networks are widely used in this problem. However, classical network also face some problems such as overfitting and hard to determine the parameters. Here we use a multilayer feedforward neural network, designed by Monte-Carlo-Adaptation, to deal with the index futures’s time series prediction problems. The result show that, MCA network can constrain the sensitivity of network inorder to suppresses ovefitting problem, and we can obtain a higher generalization ability.This paper is divided into three parts:The first part is a brief introduction of index futures trading characteristics and some traditional analysis methods, these methods often affect the behavior of the traders in turn; the second part present the construction of artificial neural networks, difference between networks is reflected in the learning rule, so we compared the BP network and MCA network in detail; In the last part, we use MCA network to predict the index futures’s time series, and get a better result than BP network. At the end of the paper, we discuss the deficiencies of the modle and the improvement of future works.