A Recommendation Algorithm Adapted To Dynamic Social Networks

With the rapid development of online commerce, the type and quantity of goods that users can select are also exploding. In order to recommend the products from hundreds of millions of alternatives automatically, some researches are carried out.The tradional collaborative filtering algorithms are widely used in most of commercial systems. However, the traditional collaborative filtering algorithms need to build a huge scoring matrix, which brings the cold start, high complexity and other issues. Meanwhile, with the rapid rise of social networks, users provide a lot of personal information and data on social networks. The traditional recommendation systems have not token advantage of this data and lose much useful information. This paper describes the drawbacks of the traditional recommendation system to study the characteristics of social networks, focusing on how the social network and its dynamic change information into personalized recommendation system to achieve and adaptation of the user’s current interest in the field of high-quality recommendation. The main contribution of this paper is as follows.1.Propose a classification algorithm for user groups detection.The "strength of weak ties" theory shows that online social networks can often unidirectional behavior attention to reality than friends, colleagues and other bi-social behavior to better reflect the user’s interest distribution. Based on SWT theory, This paper using the latent factor model to dig the information hidden behind the social networks and to assess the effectiveness of the algorithm by precision and recall rate of the actual classification.2.Introduce the dynamic social network information into personalized recommendation model. This paper propose a personalized recommendation algorithm based on time perception. The algorithm considers the short-term interests of users is larger than the long-term impact on their actual interest in buying behavior, social behavior by tracking the user to tap the recent short-term interests of users. The algorithm is designed to specifically quantify the impact attenuation formula of user interest change their purchasing behavior, the introduction of this impact probability matrix factorization recommended model. Experimental results surface, based on a social network dynamic personalized recommendation system can improve the quality of recommendation.Based on the above two points, the paper assesses the effect of population classification with precision and recall. The experiments show that the algorithm proposed in this paper improve the shortcomings of traditional recommendation system both in the quality and efficiency. Finally, this paper proposes a recommendation algorithm based on dynamic social networks to better meet users’ demand for personalized recommendation service.