The Research On Collaborative Filtering-based Personalized Recommendation Algorithm And System Implementation

With the rapid development of the Internet,the amount of information is growing rapidly and massive data problems become increasingly serious,which makes it difficult for users to find the information of interest quickly.In addition to the popularity of mobile devices,users are more willing to find information on them.In such circumstance,personalized recommendation system came into being.Motivated by the practical need,the author learns the theoretical knowledge of the recommendation system and studies the relevant recommendation algorithm.The collaborative filtering recommendation algorithm and the k-means clustering algorithm are emphatically studied,and the collaborative filtering algorithm is improved.In the end,the recommendation subsystem is designed and implemented.The main work of this thesis has the following aspects:1.The influence of the popular object on the user' similarity in the recommendation algorithm is analyzed,and the penalty factor is added to calculate the user' similarity to reduce the influence of the popular object on the user' similarity.Through the experiment,the accuracy and recall rate of the proposed algorithm are improved.2.For the time bottleneck and expansion problem of collaborative filtering algorithm,a clustering algorithm is proposed to heighten the collaborative filtering algorithm.Not only the users' rating information but also the feature information of rating object is used to create model and cluster for users.Through the experiment,the recommendation efficiency and the prediction accuracy of the proposed algorithm are improved.On this basis,penalty factor and users' clustering are used to improve user-based collaborative filtering algorithm.The experiment validates that comprehensive improvement is superior to any single enhancement in the recommendation efficiency and recommendation quality.3.According to the specific mobile application scenarios and specific needs of a company's business,the proposed collaborative filtering algorithm is used to design and implement the article recommendation subsystem.The system includes four modules which are collecting and processing of users' data,modeling of users and recommended objects,recommendation algorithm and the display of recommendation list.