Design And Implementation Of A Museum Recommendation System Based On Collaborative Filtering

With the popularization of the Internet and the improvement of information technology,people's lifestyles have changed greatly.People have transformed from material pursuits to deeper spiritual pursuits,and museums are playing an increasingly important role as a window for cultural transmission.In recent years,in the process of informatization construction,the museum pays more attention to the display and dissemination of cultural relics,and uses various informatization methods to display the cultural relics in the collection to the audience,but this brings about the problem of information overload.There are too many cultural relics displayed online,and users cannot see the cultural relic information they want when browsing,and the recommendation system can solve the problem of information overload to a certain extent.In this paper,the Hebei Museum Smart Museum is used as an application scenario.The existing problems in the traditional collaborative filtering recommendation algorithm are improved,and a collaborative filtering-based museum recommendation system is designed and implemented.When a user accesses a cultural relic in the system,the system recommends other cultural relic information similar to the current cultural relic to the user according to the characteristic attributes of the current cultural relic and the historical behavior data of the user.Collaborative filtering recommendation algorithms are widely used in large websites such as Amazon,Dangdang,and Taobao,and have good recommendation results^[1].However,problems such as cold start and sparse scoring matrices still exist.This article combines the application scenarios to improve the collaborative filtering algorithm as follows:1)In the traditional collaborative filtering recommendation algorithm,when calculating the similarity of two cultural relics,it is mainly considered that the two cultural relics have a common score,but the system's scoring matrix is too sparse,the error in calculating the similarity of cultural relics is large.This article considers the case where there is only one cultural relic.For cultural relics without ratings,we need to predict the user's preference for this cultural relic.This preference may be related to the user's activity and the popularity of the cultural relic.This paper comprehensively considers the influence of the two,and proposes to add related penalty factors to improve the accuracy of recommendation.2)It is proposed to use the similarity of cultural relic attributes to solve the cold start problem.If there is no user's historical score,using the unique attribute characteristics of the cultural relics to calculate the similarity with other cultural relics can improve the accuracy of the recommendation.3)Using the characteristics of artificial neural network self-learning,when calculating the similarity of cultural relics,the similarity of the historical score of the cultural relics and the similarity of the attributes of the cultural relics are used as the input of the BP neural model.After the forward propagation process,the similarity of the final cultural relics is obtained,and the predicted score of the cultural relics is calculated.Then,the error between the prediction score and the true score is adjusted in the reverse direction,and the weights and offsets are continuously updated to obtain the final prediction model.