摘要
针对网络舆情时间序列具有样本数少、非线性、贫信息等特点,本文采用改进后的混合蛙跳算法——最小二乘支持向量机模型进行网络舆情预测。首先利用反向学习策略构造初始化种群,其次利用自适应移动因子改进蛙群个体更新步长,然后根据适应度方差动态调整蛙群个体的变异概率更新全局最优解。经过改进后的混合蛙跳算法对最小二乘支持向量机的两个重要参数——核函数的宽度参数σ、正则化参数γ进行寻优,应用到网络舆情预测中。
The improved shuffled frog leaping algorithm -the least squares support vector machine （LSSVM） prediction model is used to predict the network public opinion which has the characteristics of small sample number, nonlinearity and poor information, etc. Firstly, the initial population is constructed by using the reverse learning strategy. Secondly, the individual step size of the frog popula- tion is improved by using the adaptive moving factor. Then we can adjust the mutation probability dynamically according to the variance of fitness and update the global optimal solution. The modified shuffled frog leaping algorithm is used to optimize the width parameters and regularization parameters of the two important parameters of LSSVM, and it is used to the Internet public opinion forecasting.
出处
《网络新媒体技术》
2017年第5期29-35,共7页
Network New Media Technology
基金
中国航天科技集团公司舆情监测信息化项目
关键词
网络舆情预测
混合蛙跳算法
最小二乘支持向量机回归
Internet public opinion forecasting
shuffled frog leaping algorithm
least squares support vector machine regression
