Personalized tourism recommendation algorithm integrating tag and emotional polarity analysis

Abstract

The matrix factorization algorithm has achieved good results in the field of scenic spot recommendation, but there are still many problems. Most of the studies only refer to the user’ s score of the tourist attraction to represent the user ’s overall evaluation of the attraction, thus ignoring the correlation between the online review text information of the attraction and the attraction, which eventually leads to the low accuracy of the entire recommendation system. This paper proposes a personalized tourism recommendation algorithm that integrates scenic spot labels and sentiment polarity analysis. On the basis of constructing feature labels for various scenic spots, the sentiment polarity analysis of user comments is carried out through natural language text information processing, and implicit emotions are given to the labels contained in scenic spots. Secondly, it is introduced into the factor vector of matrix decomposition model algorithm, and finally the potential relationship between users and feature labels and scenic spots is deeply explored. Through the user ’s preference for emotional feature labels, the user ’s approximate rating value for attractions is predicted. Experiments show that compared with the user-based collaborative filtering scenic spot recommendation algorithm, the average absolute error (MAE) and root mean square error (RMSE) are reduced by 70.28 % and 65.23 % respectively. Compared with the tag-based collaborative filtering attraction recommendation algorithm, MAE and RMSE are reduced by 65.02 % and 63.93 %, respectively. Compared with the matrix factorization recommendation algorithm of scenic spot labels, MAE and RMSE are reduced by 34.02% and 29.93 % respectively. To sum up the performance of the algorithm is significantly higher than the existing attractions recommended algorithm can be more accurate to provide users with recommended options.