Research On User Dynamic Preference Based Recommendation Algorithm

The tide of the Internet era has brought the explosive growth of data,thus the cost of users accessing to useful information from the massive data is higher and higher.Recommender system can alleviate this problem.Recommender system models the user's preferences by user portraits,user's historical behavior data,etc.,thus help users to find the information they are interested in quickly.At present,non-time-aware recommendation algorithm usually considers the user's preference has a static pattern.Under this assumption,the related algorithm is studied.However,in the real world,user preferences are changing over time.Current time-aware recommendation algorithm does not consider long-term,repetitive preferences of users,and the algorithms' is inefficient and so on.Therefore,the study of the user's dynamic preference to enhance the accuracy and the recall rate of the personalized recommendation algorithm is very important.In this paper,we first analyze user dynamic preferences in the e-commerce scenario,then focuses on the heterogeneous implicit feedback recommendation algorithm based on user's dynamic preference.The main work are listed as follows:1)A time decay recommendation algorithm based on user preference confidenceThe time-aware recommendation algorithm usually uses the time decay function to reduce the user's historical rating to predict the user's future purchase interest.In addition,the existing research mainly uses time attenuation in a relatively simple user-based collaborative filtering algorithm,and has not been applied to the model-based recommendation algorithm.To address the above two problems,this paper proposes a time decay recommendation algorithm based on user preference confidence.We believe that the it is user's confidence of rating,not user's rating changing over time,thus it characterizes the user's short-term preferences.The experimental results show that the proposed method can better represent the user's preference and improve the accuracy and recall measure.2)A user dynamic preference recommendation algorithm based on Hidden Markov ModelModeling user's long-term,repetitive preferences plays an important role in improving the accuracy of the personalized recommendation system.The above model based on time decay can identify the short-term preference of the user,but it is not enough to identify the user's long-term,repeatable preferences.To solve these problems,this paper proposes a user dynamic preference recommendation algorithm based on Hidden Markov Model.The method establishes a hidden Markov model for each user based on users' historical behavior data to predict the long-term and repetitive preference of the user.The results show that the proposed method can identify the long-term and repetitive preferences of users,and improve the accuracy and recall measure.