The research of margin setting model based on improved BP neural network technology

The margin system is the exchange´s first line defense against default risk. The margin setting model of stock index futures established adopts improved BP neural network technology, which utilizes Genetic Algorithm (GA) to optimize the weight of BP neural network to expand the search space and enhance learning efficiency and accuracy of the network. Taking Hong Kong Hang Seng Index Future as the research object, and selecting prudent index (PI) and opportunity cost index (OCI) to compare and evaluate the predicted effect with traditional methods EWMA model and EVT-VaR method. Empirical study result shows that the prudent index of this model is the highest and the opportunity cost is relatively low, therefore, the model proposed has a better prediction effect.