基于混合算法和BP神经网络的网络舆情预测研究

Research on the Prediction of Network Public Opinion Based on Hybrid Algorithm and BP Neural Network

【目的/意义】网民对社会现象及问题表达意见、态度使得网络舆情对社会的影响力越来越大,构建模型对网络舆情的发展进行预测具有现实意义。【方法/过程】通过信息熵理论控制种群初始化,利用遗传算法较强的全局搜索能力和粒子群算法的局部搜索能力实现对BP神经网络权值的优化,构建混合算法优化的BP神经网络的网络舆情预测模型并进行实证分析及对比实验。【结果/结论】结果表明,该模型在预测性能上具有更好的优越性及稳定性。

[ Purpose/significance ] The attitudes and opiniuns t^wards social phenomena and issues which expressed by In- ternet users are making more and more impact on society. Therefore, it is of practical significance to construct the model to predict the development of public opinion. [ Method/process ] Controlling the population initialization by information entropy theory. And then, optimizing the weight of BP neural network by using the strong global search ability of genetic algorithm and local search ability of particle swarm algorithm. Construct the prediction model based on BP neural network that optimized by hybrid algorithm to carry out empirical analysis and comparative experiments. [Result/conclusion] The results show that the model has better superiority and stability in prediction performance.