Analysis Of Tourists’ Online Comments Based On CNN-BiLSTM Combination Model

With the continuous development of economy and consumption level,tourists have higher requirements for tourism services.In order to better meet the demands of tourists,analyzing tourists’ travel preferences based on big data is of great significance for improving scenic spot management,accurately locating tourists’ needs,and optimizing scenic services.Taking Hunan popular scenic spots as an example,based on the online comments data of tourists in Hunan popular scenic spots on Trip.com Group website,this paper uses the deep learning model to conduct sentiment analysis of tourists’ online comment data.Firstly,for the online comment data of tourists,the word vector is trained by word2 vec method,and the CNN-Bi LSTM combination model is constructed to classify the sentiment of the comment data.Based on the experimental results,it is found that compared with Text CNN,RNN,LSTM,Bi LSTM and other models,the comprehensive evaluation index F-Measure of the CNN-Bi LSTM combined model has increased by about 1.7%,2.5%,1.0% and 0.4% respectively,indicating that the model has a better effect on the sentiment classification task of tourists’ online comments.Secondly,the LDA theme model is used to analyze the main reasons for tourists’ favorable and unfavorable reviews.Through the in-depth analysis of the results,it can be known that the factors affecting the favorable comments of scenic spots mainly include cultural and historical heritage,night scenery,and related services of scenic spots,while the factors affecting the negative evaluation mainly include queuing time,ticket price,scenic spot business,weather conditions,etc.Finally,combined with the actual situation of tourism,aiming at improving the quality of scenic spots and tourists’ tourism experience,we give some reasonable suggestions for the development of scenic spots.