基于模糊粒化的改进混合神经网络股指期货价格区间预测

发布时间：2018-03-24 07:12

本文选题：股指期货　切入点：模糊信息粒化　出处：《南方金融》2017年11期

【摘要】：为提高区间预测的精度,提出一种基于三角模糊信息粒化的改进径向基(RBF)与支持向量回归机(SVR)相结合的混合神经网络区间预测模型,对股指期货价格的变化区间进行预测。首先,对原始数据进行模糊粒化处理,获得相应的变化区间;其次,采取自组织学习策略并运用减聚类算法,对传统的RBF神经网络进行优化,改进模型的结构与参数;然后,运用SVR对模型滚动预测过程中产生的残差趋势作进一步的估计,从而修正预测值;最后,运用改进混合神经网络对模糊粒化后的沪深300股指期货数据进行实例验证。结果表明,基于模糊信息粒化的改进混合神经网络区间预测模型能够较为精确地预测股指期货价格的变化范围与价格走势,有效提高单一非参数模型的点预测与区间预测的精度和运行效率,同时具备较好的网络结构与拟合能力。
[Abstract]:In order to improve the accuracy of interval prediction, a hybrid neural network interval prediction model based on triangular fuzzy information granulation (RBF) and support vector regression machine (SVR) is proposed. The change interval of stock index futures price is forecasted. Firstly, the original data is processed by fuzzy granulation, and the corresponding change interval is obtained. Secondly, self-organizing learning strategy is adopted and clustering algorithm is used to reduce the value of stock index futures. The traditional RBF neural network is optimized to improve the structure and parameters of the model. Then, SVR is used to further estimate the trend of residual error in the process of model rolling prediction, so that the prediction value can be revised. The improved hybrid neural network is used to verify the fuzzy-granulated data of Shanghai and Shenzhen 300 stock index futures. The results show that, The improved hybrid neural network interval prediction model based on fuzzy information granulation can accurately predict the range and price trend of stock index futures price. The accuracy and operation efficiency of point prediction and interval prediction of single nonparametric model are improved effectively, and the network structure and fitting ability are also improved.
【作者单位】：华南理工大学工商管理学院;
【基金】：国家自然科学基金项目《区域孵育网络体系建设运营评价及其风险度量》(项目编号:71073056) 教育部人文社会科学项目《P2P的信用风险评价与监管研究》(项目编号:Y9150040) 中央高校重大决策项目《互联网金融与孵化器企业集群融资创新及其对策研究》(项目编号:D214433W)的资助
【分类号】：F724.5;TP183

【相似文献】