A Hybrid News Recommendation Based On Time Factor

The rapid development of the Internet has led to the rapid growth of information on the Internet.Human society has changed from a lack of information to an information overload stage.Nowadays,people can obtain a large amount of data through various channels.This indirectly leads to the daily exposure of Internet users.The quality of information is uneven.If users are completely screened by themselves,they will waste too much time and energy and seriously affect the user's reading experience.At the same time,users are no longer satisfied with viewing information that satisfies popular reading interests,but also prefer to be able to browse customized information that matches their reading interests.The personalized recommendation system can help the user to filter out the information content that meets the user's reading interest from the mass data,so that the user can save a great deal of reading costs.Selecting the recommendation algorithm that meets the application scenario generates content that suits their reading habits for the user,and can discover potential reading interests that the user has not yet discovered,and expands the diversity of user reading information.However,the traditional personalized recommendation can not efficiently and reasonably generate a recommendation list for the user.It also needs to analyze the user's historical behavior data to dig out more user characteristic behaviors,study the regularity of the user behavior,and discover the potential reading interests of the user.This will more reasonably predict the user's future reading interests and generate content more in line with the user's interest in reading.By analyzing the daily behavior of users through big data research,we can find that the user's daily schedule has a certain regularity,and the daily news reading duration does not change much,and it is possible to study the potential relevance of users' news reading interest changes and schedules,and to generate more information.Reading interest model for reading scenes.Therefore,this paper proposes a hybrid recommendation algorithm based on time factor.The main contents of the algorithm are:(1)Collecting historical reading data of users,studying their reading interestchanges at different times,analyzing the user's reading needs at different times,and providing recommendations that are more consistent with the user's reading habits.(2)A new hybrid recommendation algorithm is proposed.According to the user's reading behavior is active reading or passive reading,combined with the user's long-term,short-term reading interest,according to different application scenarios,a more reasonable recommendation algorithm is used to generate the recommendation list.(3)In order to improve the clustering effect and reduce the impact of noise behavior data on the clustering results,this paper selects DBSCAN clustering algorithm.At the same time,in order to reduce the impact of parameters Eps and minPts on the final result in the DBSCAN algorithm,an improved algorithm I-DBSCAN was chosen.(4)In order to generate the vector space model of text better,the TF-IDF algorithm is used to weigh the importance of each keyword in the news content to be recommended,and the commonly used vocabularies with little relevance to the topic of the text are filtered out,and then the remaining words are The important vocabulary uses the LDA algorithm to determine the probabilistic subject of the text,ensuring that the calculation results are more likely to reflect the text theme.