Sentiment Classification And Theme Mining Of The Hotel Review

Nowadays,the rapid development of the Internet provides great convenience for people's lives.The reviews can bring key information to consumers' decisions.Consumers can use network resources to meet their own needs.Tourists can complete the hotel reservation through the online platform,where they can consider the visitors' reviews of the hotel's accommodation as a reference.At the same time,the hotel will also find its own advantages and disadvantages according to the comments of tourists.However,the quality of comment information is uneven and the amount of data is very large.It is difficult to understand the specific situation of the hotel through manual preview.Therefore,it is very meaningful to use the classical machine learning method to classify the hotel reviews accurately and to explore the potential information.First of all,this thesis uses data from Ctrip.The data is preprocessed by cleaning,Chinese word segmentation and so on.In order to interpret the data more intuitively,this thesis makes a descriptive statistical analysis of the preprocessed data.There were 5322 positive reviews and 2444 negative reviews.It can be seen that the positive reviews of the hotel outweigh the negative reviews and the hotel's review tendency is positive.After that,in order to judge the emotional tendency of hotel reviews,five text classifiers are constructed for pre-processed data,namely,logical regression,naive Bayes,support vector machine,random forest,XGBoost,etc.And the accuracy rate,recall rate,F1 value will be considered as a measure of the model.The experimental result show that logical regression has the largest F1 and AUC,so it has the best classification effect on the test set.Therefore,the logical regression model is selected to realize the classification of text.Finally,the LDA theme model is constructed for two types of reviews,and the potential themes are explored.Thus we can find opinions on the hotel.The advantages of hotel are excavated from positive reviews,and the shortcomings are excavated in negative reviews.And according to the potential information are excavated above,I will put forward targeted suggestions for the hotel to promote the development of the hotel.