Design And Implementation Of Real-Time Personalized Recommendation System

With the rapid development of Internet technology, we can reach more and more information, but the rapid growth of information leads to a reduction in the utilization of information. Personalized recommendation system is an intelligent system in order to solve this problem. In recent years, with the widely using in daily life, personalized recommendation system has improved rapidly. But in most of the traditional personalized recommendation systems, they analysis the data regularly, then update the recommend mode, finally make a recommendation based on the new mode. Because of the regularly updating of the mode, the recommendation result may not reflect the current state, which may lead to an inaccurate result。In this paper, I designed a real-time personalized recommendation system which updates the mode and make a recommendation in real-time.Based on the in-depth study of traditional recommendation algorithm, combined with the distributed stream processing technology, I designed and realized a real-time personalized recommendation system, and researched the real-time recommendation system from real-time trend recommendation and real-time items recommendation.Real-time items recommendation, namely according to the real-time behavior of the users, recommend similar items, mainly used in electronic commerce platform. The main research contents include:(1)According to the characteristics of real-time items recommendation, I designed the system architecture which combines off-line processing with online processing. This architecture can make full use of the traditional off-line recommendation algorithm, and combined the online incremental updating algorithm, improve the timeliness and accuracy of recommendation. (2)In order to improve the accuracy of recommendation system, I designed and realized real-time incremental updating algorithm using the data steam processing technology, which can update the recommendation model according to the real-time action, improve recommendation accuracy, and has good scalability.(3)To calculate the recommendation result in a short time, I designed and realized the online recommendation algorithm, which calculate the recommendation result using the off-line processing result and actions of current period.Real-time trend recommendation system, namely statistics the real-time trend according to the action in current period, which mainly used in hot topic in social network. The main research contents include:(1)In order to ensure the timeliness of the trend recommendation, I combined the real-time data statistics with sliding windows, and used Storm to realized the real-time trend recommendation, which can statistics and update the real-time trend according to the change of the real-time data.(2)To reduce the coupling of the data sources and real-time processing logic, I designed and realized the processing logic using Message Queue.Finally I realized the real-time item recommendation and the real-time trend recommendation in practice and designed some experiments to verify the improvement of the accuracy. To meet the requirement of the real-time in recommendation system, we design and construct a real-time personalized recommendation system, which consists of real-time item recommendation function and real-time trend recommendation function. We design some experiments to verify the feasibility of the system.