摘要
针对网络舆情数据随机波动大、原始样本少的特点,对网络舆情短期预测方法开展研究。基于混沌理论对原始样本进行相空间重构,确定了最佳延迟时间和嵌入维数。采用LS-SVM对网络舆情数据进行回归建模,基于粒子群算法对LS-SVM参数进行优化,避免了核函数参数选择的主观性和盲目性。以某网络事件点击量预测为案例进行了仿真实验。结果表明:所提方法具有预测精度高、能确定最佳模型参数的优点,从而验证了所提方法的科学性和先进性。
Regarding the characteristics of little original sample and big fluctuation of network public opinion data,a prediction method of short-term network public opinion was studied.The phase space of original sample was reconstructed based on chaos theory.The optimum delay time and embedding dimension were acquired.The regression model of network public opinion was built by LS-SVM.The LS-SVM parameters were optimized based on PSO,which can avoid subjectivity and blindness of choosing kernel function parameters.Taking prediction of a network event’s clicks as an example,simulation was done.Simulation results show that the method proposed above has high prediction precision and can get optimum model parameters.So the method proposed above is scientific and advanced.
作者
高颖
GAO Ying(Zhou Enlai School of Government,Nankai University,Tianjin 300000,China;Personnel Department,Inner Mongolia University for the Nationalities,Tongliao 028000,China)
出处
《重庆理工大学学报(自然科学)》
CAS
北大核心
2019年第6期171-176,共6页
Journal of Chongqing University of Technology:Natural Science
基金
教育部人文社科项目(14XJC840002)
内蒙古民族大学科学研究基金资助项目(NMDYB18043)