Text Mining And Comprehensive Ranking Of Hotels Based On Hotel Reviews

In the era of rapid development of the Internet,users express their opinions and views on products or services in the form of online comments,thus forming a large amount of comment text information.These textual information have potential value for merchants and consumers.On the one hand,merchants can learn about customers’ attitudes towards certain products and services and their specific evaluations in certain aspects through textual information.On the other hand,customers can browse online reviews to compare multiple similar products to make a choice.In a sense,online review information affects the decision-making process of potential consumers to a large extent,and can also provide reference for the improvement of products or services.However,in the face of cumbersome review text data,how to mine objective knowledge about products from it,and how to make final decisions from alternative products based on text information deserves further study.To this end,based on the hotel review text,this paper completes the following work for these two questions:First of all,this article uses the Octopus crawler tool to obtain relevant information about the reviews of Ctrip users on Hanting,Jinjiang Inn,Home Inn and7 Days and other similar budget hotels in Zhengzhou.The original corpus data set is cleaned,and the preprocessing of the initial data set is completed through the steps of text deduplication,mechanical compression,short sentence deletion,word segmentation and removal of stop words.Secondly,for a hotel,taking Hanting as an example,the method of sentiment analysis and topic model is used to realize the text mining of the hotel reviews.Specifically,the sentiment classification model is trained by machine learning classification algorithms such as Naive Bayes,Decision Tree,Random Forest,SVM,Logistic Regression,XGBoost,and deep learning algorithms such as LSTM and BERT to effectively identify the sentiment tendency of comment texts.And use LDA topic modeling to mine positive reviews and negative reviews respectively,and then explore the factors that customers are satisfied or dissatisfied with the hotel.Finally,for multiple hotels,the evaluation attributes and the weight of each attribute are determined based on the review text information,an intuitive fuzzy multi-attribute evaluation matrix is constructed,the positive and negative ideal values of each attribute are found,and the comprehensive evaluation value of each candidate hotel is obtained by using the TOPSIS method.Thereby,multiple different hotels can be comprehensively sorted to help customers achieve the best decision.