主成分分析与BP神经网络在微博舆情预判中的应用

Application of Micro-blog Public Opinion Prediction Based on Combination of Principal Component Analysis and BP Neural Network

首先通过主成分分析消除原始指标之间的相关性,使指标数量变少且相互之间不相关,从而构建综合预判指标,再利用BP神经网络建立微博舆情预判模型。实验选取2013年微博热门话题作为训练样本,选取2014年的话题作为预测。实验结果表明,主成分分析有助于去除原始样本数据的冗余,简化了网络的复杂度,所得到的结果更加准确。因此,该模型较仅使用BP神经网络的准确性更高。

Firstly, the paper reduced the correlation of original indicators by principal component analysis and the number of indicators and made them separate from each other. scondly, it established comprehensive indicators for prediction. BP neu- ral network is used to establish the model of micro - blog public opinion prediction. The experiment using 2013 micro - blog hot topics to predict 2014 hot topics. Experimental results showed that principal component analysis helped reduce redundancy of the statistical data and made predictions more accurate. This model was more accurate than the model using only BP neural network.