Design And Implementation Of Personalized Novel Recommendation System

With the rapid spread of online literature in the Internet,more and more authors have signed up for small novel websites to seek development chance.However,small novel websites use traditional recommendation backgrounds,and the recommendation ability is poor and it is not easy to iteratively upgrade.It is impossible to effectively mine high-quality works to attract readers.Therefore,the novel website recommendation system is connected to the novel website,which can personalize the recommendation,quickly expand the recommendation strategy,and can verify the recommendation effect.This paper designs and implements the personalized novel recommendation system,and adopts a distributed architecture to ensure the availability,consistency,robustness and scalability of the system.After a comprehensive analysis of the system,it is divided into four modules:online recommendation engine module,reading traffic scheduling module,offline computing module,and cold start rating module.The author independently designed and implemented the above four modules:(1)Online recommendation engine module:This module can personalize online recommendation novels to readers.Operators and developers can configure recommendation strategies and monitor engine performance troubleshooting systems in real time.(2)Reading traffic scheduling module:This module can schedule reading traffic and find the optimal recommendation strategy.The operator configures the traffic scheduling policy.The module groups the reading traffic online,and uses different recommendation strategies for each group.After that,the operating personnel compare the recommended results to obtain the optimal recommendation strategy.(3)Offline calculation module:This module can calculate novels that may be of interest to readers.An item-based collaborative filtering algorithm is implemented,which reads the reader's reading data daily,and uses the big data computing framework to calculate the similar novel collection of the novel offline.(4)Cold start rating module:This module can provide readers who do not have reading records to provide recommended reading novels.The novel rating task is executed regularly every day,and the cold start recommendation list is arranged according to the level.The results of the functional test,stress test and A/B test show that the personalized novel recommendation system fully meets the user's requirements.With the expansion and management of the recommendation strategy in the later period,the system will be able to further enhance the recommendation effect,use the reading traffic more effectively,explore more quality works and create more commercial value.