Design And Implementation Of Collaborative Filtering Recommendation Based On Group Discovery And Interval Division System

With the rapid spread of the Internet,various industries have entered the Internet era.The rapid growth of the number of users and the number of industry projects has brought about problems of information overload,and massive amounts of data have made it difficult for users to easily obtain information of their own concern.Following the search engine,the emergence of personalized recommendation technology frees the user from inputting descriptive information.As an important method to solve the problem of information overload,personalized recommendation technology has played an important role in e-commerce,video,music,social networking and other fields.Among them,the collaborative filtering algorithm is the most widely used,most successful and has been a lot of research and use.Collaborative filtering algorithm has intuitive flow,but in the information explosion environment,problems such as expansibility and sparse data are exposed.These problems often greatly affect the recommendation efficiency and accuracy of recommendations.For sparsity and recommendation accuracy issues,this article develops the following studies:Firstly,aiming at the problem of data sparsity in traditional collaborative filtering,a group preprocessing method based on spectral clustering and FMC clustering is proposed.The algorithm converts the user project scoring matrix into undirected graphs,then combine similarity calculation and spectral clustering algorithm to obtain groups in the form of feature vectors,the FMC initial clustering center is optimized by combining the maximum and minimum similarity methods,and the membership degree of the user and the project to the group is calculated by the eigenvector clustering,and the membership matrix of the group and the user project to the group is finally obtained The experimental data was designed in combination with standard data sets.Experiments show that this algorithm can effectively reduce the sparseness of data and can improve recommendation accuracy when combined with recommendation algorithms.Second,on the basis of solving the sparsity problem in the previous step,aiming at the problem that the traditional similarity algorithm ignores the difference of user rating criteria,a collaborative filtering algorithm based on the user's preference interval division is proposed.The algorithm counts and calculates the user's probability of using the score value,maps the probability and divides the user's preference interval,converts the preference interval into a coordinate form,and combines similarity calculation methods with the nearest neighbor set,predicting the target user by the nearest neighbor's score,then scores and combine groups to give final recommendations.The experimental data was designed in combination with standard data sets.Experiments show that this algorithm can significantly improve the accuracy of recommendations.The above algorithm is applied to the IKEA project recommendation module,and the main nodejs function library,recommended architecture design,and implementation details of the core part are introduced in detail.First,the group was mined using pre-processing method,and the collaborative filtering algorithm based on the user's preference interval was directly used in the group,and combine all groups to achieve the final recommendation.,and a good recommendation effect was obtained.