基于加权Word2Vec-XGBoost的情感分析方法研究  

Research on Sentiment Analysis Method Based on Weighted Word2Vec-XGBoost

作　　者：王玉珍 王彤 WANG Yuzhen;WANG Tong(School of Information Engineering,Lanzhou University of Finance and Economics,Lanzhou 730020,China;Key Laboratory of E-commerce Technology and Application of Gansu Province,Lanzhou University of Finance and Economics,Lanzhou 730020,China)

机构地区：[1]兰州财经大学信息工程学院,兰州730020 [2]兰州财经大学甘肃省电子商务技术与应用重点实验室,兰州730020

出　　处：《重庆科技学院学报（自然科学版）》2023年第3期62-66,共5页Journal of Chongqing University of Science and Technology：Natural Sciences Edition

基　　金：甘肃省哲学社会科学规划项目“甘肃省数字经济评价体系及发展对策研究”(2022YB077);甘肃省自然科学基金项目“甘肃省数字经济发展水平评价及耦合协调分析”(22JR5RA558);甘肃省电子商务技术与应用重点实验室开放课题(2023GSDZSWA02);兰州市哲学社会科学规划项目“兰州营商环境存在的短板及解决路径研究”(22-A09)。

摘　　要：针对网络评论信息数据不平衡问题,提出了基于加权Word2Vec-XGBoost的情感分析方法,以提高情感分析的准确率。首先,利用TF-IDF算法对Word2Vec词向量模型进行加权,并将其作为模型输入;然后,利用XGBoost模型进行情感分析。实验结果表明,加权Word2Vec-XGBoost模型比传统的逻辑回归和决策树分类模型的情感分类准确率更高。Addressing the issue of imbalanced data in online coment information,a sentiment analysis method based on weighted Word2Vec model and XGBoost model is proposed to improve the accuracy of emotional analysis.Firstly,the TF-IDF algorithm is used to weight the Word2Vec word vector model as the model input,and then the XGBoost model is used for sentiment analysis.The experimental results show that the weighted Word2Vec-XGBoost model is more accurate than the traditional logistic regression and decision tree classification models in sentiment classification.

关 键 词：情感分析 Word2Vec TF-IDF XGBoost 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]