Research On Personalized Tourism Recommendation Algorithm Integrating Tag And Emotional Polarity Analysis

In recent years,the ideology of the Internet has undergone tremendous changes.In the past,we can obtain all kinds of information through active search,but with the passage of time,the accumulation of more and more redundant information has caused serious information overload to the whole search engine,and the traditional search engine model is no longer applicable to the current development form.With the popularity of artificial intelligence and machine learning models,the method of obtaining information has changed from the initial online convenient query to personalized recommendation.The purpose of the recommendation system is to actively understand the user ’s heart and mind,and then make selective strategy recommendations to the user.It has been applied in various fields of life.Among them,personalized tourism recommendation has gradually emerged in recent years,but there are many problems.First of all,the data set in this field is scarce compared with other fields,and there is no public data set suitable for research.Secondly,the research related to the algorithm in the field of tourism recommendation is relatively rare,and the accuracy of strategy recommendation is low.Finally,in order to protect the user ’s personal privacy,it is often only recommended by objective user ratings when making travel recommendations,resulting in a failure to truly reflect the user ’s true inner portrayal of the attractions.In the face of the above problems,this thesis constructs a new personalized scenic spot recommendation model algorithm.Firstly,on the data set,this thesis constructs the data set of this recommendation model by mining several large tourism platforms.Secondly,in terms of user data,this thesis uses the polar sentiment semantic analysis method of natural language processing in the user ’s text comment data for the attractions,and gives a user a subjective emotional evaluation of the attractions,which increases the interpretability of the user ’s real evaluation of the attractions.Next,in order to improve the accuracy of the whole tourism recommendation model,this thesis proposes a personalized matrix factorization algorithm that integrates label and sentiment polarity analysis.This method not only establishes the scenic spot label of specific attributes for all scenic spots.Based on the matrix factorization algorithm model,in order to take into account the limitations and inaccuracies of the actual objective score,this thesis constructs a new subjective scoring matrix with the same dimension for the emotional features generated by the user ’s text review of the scenic spot,and fuses the objective actual scoring matrix with the emotional feature matrix through continuous experiments.Finally,the experiment is carried out on the constructed data set,and the(mean absolute error,MAE)and(root mean square error,RMSE)of the recommendation model are calculated by comparing the three proposed recommendation algorithms in the tourism field.Through experiments,it is concluded that the MAE and RMSE of the model algorithm proposed in this thesis are greatly reduced.Compared with the User-CF recommendation algorithm,it is reduced by 70.28 % and 65.23 %.Compared with the Label-CF recommendation algorithm,it is reduced by 65.02 % and 63.93 %.Compared with the Label-MF recommendation algorithm,it is reduced by 34.02 %and 29.93 %.It shows that this model algorithm is superior to other proposed models in the field of tourism recommendation at this stage.The model algorithm conforms to the user ’s habits in life and has certain rationality.It provides a new idea for the application of this algorithm to other large tourism platforms or other fields.