Research And Application Of Personalized Recommendation Algorithm Based On Collaborative Filtering

The rapid development of the Internet and e-commerce has met many people's needs.People often deal with massive data in daily life.How to obtain useful information in overwhelming data is a research value.A collaborative filtering recommendation system has emerged.Collaborative filtering recommendation algorithm is the soul of collaborative filtering recommendation system.It considers the relationship between users and commodities,and actively recommends products for users,but ignores the possible connections between users,and has problems such as cold start and data sparse.In addition,the rating prediction is an important part of the recommendation system.Generally,when it recommended items to user,firstly it predict that how many points the people will give the item.The predicted score is closer to the actual score,the recommended result is better.The traditional rating prediction algorithm has a wide gap between the score and the real score.Therefore,this paper has done further research on user-based collaborative filtering algorithm and film-based rating prediction algorithm.The main research work and achievements are as follows:(1)An improved user-based collaborative filtering algorithm is proposed.The algorithm first finds the similarity between the users according to the products purchased by the user,and then calculates the similarity between the user positions according to the geographic location of the user,and then weights the two to obtain the final similarity.The final similarity is sorted by descending order,we can find the top N users according to the similarity,put them in the neighbor user set,and produce a list of recommended products for the user.The algorithm improves the accuracy and recall rate of the recommendation.When the user is new or the score data is small,the traditional collaborative filtering cannot be recommended,and the improved algorithm can recommend the item by the position.The algorithm can effectively alleviate the cold start and data sparse problems.(2)The collaborative filtering recommendation algorithm based on film rating prediction is proposed,which is an optimization of the traditional film rating prediction algorithm.The traditional film rating prediction algorithm utilizes the user's history score record for the movie,without considering the movie attribute or user attribute.We first use the attribute information of the movie to construct the user's self-portrait;then calculate the similarity between the historical score records of the users and the similarity between the user's self-portraits,and weighted the two similarities to construct a new user similarity calculation method.Finally,predict the film score.The algorithm reduces the error of the rating prediction,and its MAE value is less than 3.21,while the MAE value of the traditional rating prediction is between 3.21 and 3.43.(3)A collaborative filtering recommendation system was designed and developed to compare the recommended algorithms and display the recommended results.Based on the demand analysis of the collaborative filtering recommendation system,the collaborative filtering recommendation system is completed.The design requirements are met and the operation is normal.It can provide personalized recommendation service for users.There are main two innovations in the paper:1)improved user-based collaborative recommendation algorithm can alleviate data sparseness and cold start problems;2)A proposed method for determining weights is proposed in collaborative recommendation algorithm based on film rating prediction.The predicted value of this method is closer to the actual score than the traditional predicted value.