Research On Emotional Analysis Of Rental Software Comments Based On Machine Learning

With the development of urbanization and rising house prices,more and more people choose to rent houses to solve their housing problems.With the development of the Internet and the popularity of smart phones,more and more rental software appeared in people’s vision,and the house rental business began to change from offline to online.All major app stores have collected a large number of software user comments.These comment texts contain a lot of emotions and information.Analyzing these emotions and information has rich commercial value and practical significance.This thesis is based on rental software review text for emotional analysis.Firstly,the user comment data is preprocessed,including data cleaning,text word segmentation and de-stop words.And the comment text is labeled based on the combination of emotional dictionary and manual labeling.Word2 vec is used to get the word vector representation of the comment text.Secondly,the emotion tendency classification model is constructed for the preprocessed data,and the classification model based on traditional machine learning algorithm and deep learning method are constructed respectively.Then we use Support Vector Machine,Random Forest,Convolution Neural Network,Bidirectional Long-term and Short-term Memory Network,and BERT model to build and evaluate the emotional propensity classification model.Ultimately,we choose the BERT-based emotional classification model.Its accuracy rate is 89.6%,F1-socre is 0.9034,AUC value is 0.878,which is good for identifying the emotional tendencies of rental software users’ comments text.Finally,the LDA theme model is used to mine the text of rental software user reviews.The data set is divided into positive and negative emotional datasets,and the positive and negative potential topics of the users are mined separately.Based on the results of the potential topics,suggestions are provided to the rental software platform.