基于GD-FNN的金融股指预测模型

Prediction model of financial stock index based on GD-FNN

针对股票市场内部结构复杂性和外部因素多变性,构建一种基于椭圆基函数且能够动态调整网络结构的广义动态模糊神经网络模型对金融股指进行预测。以上证指数为例,在价格和成交量的基础上,将与股票市场密切相关的宏观经济指标引入模型预测指标体系。通过滑动时间窗对数据集进行处理,提高了模型预测准确性并降低了运算时间。与其他神经网络模型预测效果进行比较,结果表明提出的模型具有较好的预测效果。

Aiming at the complexity of inside structure and variability of external factors in system of stock market which make stock market prediction a complex problem,this paper proposed GD-FNN financial index prediction model which is based on elliptic basis function and could dynamically adjust the network structure. Set the predictive index system of GD-FNN which involved Shanghai composite index’s price and volume,analyzed macroeconomic indicators closely related to stock market and long-run equilibrium and cause and effect relationship among the variables. In order to get optimized result-parameters of model with fixed-length time series data sets,set a sliding window,which could improve model prediction efficiency and reduce computation time. GD-FNN shows smaller deviation and higher accuracy in prediction of Shanghai composite index,comparing with other neural network model.