基于LDA和XGBoost模型的环境公共服务微博情感分析

Microblog sentiment analysis of Jiangsu environmental public service based on LDA and XGBoost models

利用Python软件编程获取2012年1月1日至2016年12月31日间相关微博舆情数据24922条,采用LDA模型提取噪声污染、城市环境公共卫生、化工危险品、饮用水安全、水体污染、企业环境整治、环保监督与执法、空气污染8个主题,利用XGBoost集成模型对8个主题的微博舆情数据进行情感分析。分析结果表明:环保监督与执法、空气污染和噪声污染主题强度较高,均超过0.3;XGBoost模型对情感极性倾向的预测准确率达到89.6%,其中环境公共卫生、环保监督与执法、企业环境整治和空气污染呈现较显著的负向极性情感。研究结果可为江苏环境公共服务供给侧改革提供决策参考。

This research crawls 24922 pieces of microblog public opinion data by Python software from 1 st January,2012 to 31 th December,2016.Then,the LDA model is used to extract 8 topics,including noise pollution,city environmental public health,hazardous chemicals,drinking water safety,water pollution,corporate environmental governance,environmental supervising and law enforcement,and air pollution.In addition,XGBoost algorithm is used to conduct emotional analysis on the microblog data of the 8 topics.The results show that the theme intensity of noise pollution,air pollution and environmental supervising and law enforcement all exceeds 0.3,which is relatively higher.And the tendency prediction accuracy of XGBoost classifier reaches 89.6%,which means that the result of emotional analysis is more accurate.The emotional analysis results show that the environmental public health,environmental monitoring and law enforcement,corporate environmental governance and air pollution have significant negative polarity.The research results provide a decision-making reference for the supply side reform of Jiangsu environmental public service.