基于FOA-ELM的客户基金购买行为预测仿真  

Funds Exchange Prediction Based on Extreme Learning Machine and Fruit Fly of Algorithm

作　　者：李栋[1] 张文宇[1] 

机构地区：[1]西安邮电大学经济与管理学院,陕西西安710061

出　　处：《计算机仿真》2014年第6期233-237,共5页Computer Simulation

基　　金：陕西省自然科学基金(2012GQ8050);陕西省教育厅专项科研计划项目(13JK0403);西安邮电大学中青年基金(104-0410)

摘　　要：在客户基金购买行为预测问题的研究中,分析和预测客户基金购买行为对于基金销售公司而言具有非常重要的经济意义。针对客户基金购买行为预测中应用极限学习机预测存在精度不理想的问题,提出果蝇优化极限学习机预测方法。在上述方法中,果蝇优化算法不断优化极限学习机的输入层与隐含层连接权值和隐含层阈值,并以优化后的结果为基础,构建极限学习机预测模型,并将预测模型用于顾客基金交易预测。仿真结果表明,与极限学习机、灰色神经网络以及果蝇优化灰色神经网络等方法相比,改进的果蝇优化极限学习机在顾客基金交易预测中具有更高的预测精度。Analysis and prediction of customer funds purchase behavior has very important economic significance for fund sales company. Extreme Learning Machine （ELM） for customer funds purchase behavior prediction has accuracy problem. A prediction method based on ELM and fruit fly of algorithm （FOA） was proposed. In the method, the ELM input weights and thresholds of hidden layer were optimized by FOA. Then, the ELM prediction model was established by the optimized results. Finally, the prediction method was used to customer fund transactions prediction. Compared with ELM, Grey Neural Network, and FOA - GNN, FOA - ELM has better prediction accuracy.

关 键 词：基金 购买行为 预测 果蝇优化算法 极限学习机 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]