基于极限学习机的上证指数预测与分析

针对证券指数具有随机性、时变、波动性较大、非线性等特点,传统线性预测方法预测精度低等缺陷,提出了一种基于极限学习机的证券指数预测方法。极限学习机克服了BP神经网络的训练速度慢、过拟合、局部极值等缺陷,具有训练速度快、全局最优和泛化能力优异等优点。采用1991~2013年上证指数对算法性能进行训练,2014年数据做测试,对100个测试数据仿真结果表明,复相关系数高达0.9935,极限学习机是一种预测精度高、误差小的证券指数预测算法,预测结果可以为用户提供有价值的参考意见。

Forecast and Analysis of Shanghai Composite Index Based on Extreme Learning Machine

For the features of stock index such as randomness , time-variant, volatile and non -linear, and for the defects of traditional linear prediction methods like low accuracy , we propose a method for forecasting stock index based on extreme learning machine.Extreme Learning Machine has the merits of high training speed , excellent global optimum and generalization ability , etc., which overcome the shortcomings of the BP neural network such as low training speed , over-fitting , local minima.Using the data of Shanghai Stock Ex-change from 1991 to 2013 on the performance of the algorithm and data of 2014 for testing , 100 test data simu-lation results show that the multiple correlation coefficient is up to 0.9935.Extreme Learning Machine is a stock index prediction algorithm with high precision and small error , which can provide a valuable reference for users.