Research On The Hybrid Recommendation Algorithm Combining Time Decay

Massive information on the Internet provides people with a rich source of information,but this also creates information overload problems.The recommendation system can actively help users obtain the desired information,which greatly facilitates people's lives.In recent years,the recommendation system has been widely used,and it has also become a research hotspot for people.Collaborative filtering algorithm is the classical algorithm of the recommendation system,and many new algorithms are also improved on this basis.The accuracy rate or recall rate of these models has been improved to varying degrees,but the improvement of a single indicator is increasingly unable to meet the needs of users.At present,many of the recommendation system researches focus on static data.However,the user's interest has a close relationship with time.In order to tap the connection between the item and the time,the user is recommended to the item,and the user's potential interest can also be found.This article focuses on issues such as accuracy,diversity,and changes in user interests.The research work includes the following aspects:(1)In order to solve the problem of the accuracy of the recommendation system and increase the diversity of the recommended articles.This paper proposes a GP hybrid recommendation model,which combines the GLSLIM model with the PLSA model by generating the recommendation list.Then the recommendation results through the two algorithms combined with appropriate weight factors to generate a new recommendation list.It can recommends the user to the desired product while also tapping the user's potential interest,so as to better meet the user's needs.(2)User's interest is changing with time.This paper improves the time function and introduces it into the scoring matrix of the GP hybrid recommendation model to form the T-GP model(the hybrid recommendation algorithm combining time decay).This not only makes it possible to recommend the user's favorite products more rationally,but also filters uninterested and outdated products to optimize the recommendation results.This article uses the published data set Movielens in the experimental part,andcompares the difference between the algorithm and the traditional algorithm by calculating the average absolute error(MAE)and recall rate(RECALL).The T-GP model uses the time factor to reduce the noise that affects the recommendation accuracy,and the mixed model improves the recommendation result.Finally,the experimental results show that the proposed algorithm is superior to the traditional algorithm in these two evaluation indicators.