基于集成算法的在线购物平台消费者评价情感分析与研究  

Sentiment Analysis and Research on Consumer Evaluation of Online Shopping Platform Based on Integrated Algorithm

作　　者：袁钰喜 陈义安[1,2] 刘晓慧 YUAN Yuxi;CHEN Yian;LIU Xiaohui(School of Mathematics and Statistics,Chongqing Technology and Business University,Chongqing 400067,China;Chongqing Key Laboratory of Economic and Social Applied Statistics,Chongqing 400067,China)

机构地区：[1]重庆工商大学数学与统计学院,重庆400067 [2]经济社会应用统计重庆市重点实验室,重庆400067

出　　处：《现代信息科技》2024年第4期101-105,共5页Modern Information Technology

摘　　要：文章对在线购物平台的消费者评价数据进行了情感分析和分类。通过使用Python实现自动化浏览器驱动和反爬虫技术,成功采集了某东购物平台的消费者评价信息。文章提出了一种改进的集成算法,将LSTM、BiGRU、BiLSTM作为分类器,分别采用Voting和Bagging方法进行集成。结果表明,与传统的贝叶斯和逻辑回归相比,LSTM+Bagging集成算法在准确率方面分别提高了5.9%和6%,而与LSTM+Voting集成算法相比,准确率提高了0.5个百分点。另外,LSTM+Bagging模型在稳定性和鲁棒性方面表现优于LSTM+Voting算法。This paper performs sentiment analysis and classification on consumer evaluation data from online shopping platforms.By using Python to realize automatic browser driving and anti-crawler technology,it successfully collects consumer evaluation information of a certain shopping platform.This paper proposes an improved integration algorithm,which uses LSTM,BiGRU and BiLSTM as classifiers,and uses Voting and Bagging methods for integration respectively.The results show that compared with the traditional Bayesian and logistic regression,the LSTM+Bagging integration algorithm improves the accuracy by 5.9%and 6%,respectively,and compared with the LSTM+Voting integration algorithm,the accuracy increases by 0.5 percentage points.In addition,the LSTM+Bagging model outperforms the LSTM+Voting algorithm in terms of stability and robustness.

关 键 词：LSTM模型 VOTING BAGGING 电商购物 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]