Scoring Prediction Model Based On Sentiment Analysis Build And Optimize

With the rapid development of tourism,more and more people are booking travel-related products online,and the evaluation of related attractions is also exploding.The scenic commentary data can not only affect the tourists' travel plans,but also help the scenic management staff to avoid weaknesses and attract Customers,and can provide data for the personalized recommendation system of the travel website.How to accurately and efficiently obtain the required data from the large-scale scenic comment data,digitize the scenic comment data,and make the score prediction based on the review text is particularly urgent.Based on sentiment analysis,this paper studies the scores of scenic commentary scores.Firstly,it introduces the basic concepts and processes of score prediction.After obtaining the data,the topic review comments data is segmented and the stop words are pre-processed,and then three types are used.Different methods are used to extract the features of the scenic comment data,and find that the appropriate feature extraction method can improve the prediction effect of the model.Then,the data is subjected to 5-fold cross-validation to avoid over-fitting of the model.Then the evaluation index used in this paper is proposed.MSE is used to compare the model adjustment and iterative process,and RMSE is used to compare the prediction effects of different models.The basic algorithm used in the scoring prediction model,the strong learner LightGBM and the main methods of model fusion are introduced in detail.Based on the basic algorithm and LightGBM,the scoring prediction model is constructed,and the prediction results are analyzed and compared.The commonly used model combination methods are introduced.These prediction models are stacked and merged.The LightGBM model is used as the secondary learning device for Stacking,and other models are used as the secondary learning device for Stacking.The primary learner,the results of the analysis found that the score results have been significantly improved.Then,some single models are combined using the averaging method as the primary learning device of Stacking fusion,and further fusion is carried out.Finally,the fusion model is used for the score prediction of scenic commentary,and the prediction result of the score prediction model is improved.