Research And Implementation Of The Collaborative Recommendation Based On The ABC Algorithm

In the context of information overload and information sharing, personalized recommender system, a new information service and filtering technology, has become a hot research topic. Especially the rapid development of e-commerce has greatly promoted the research and application of recommender system. Collaborative filtering technology evaluates and recommend items by sharing information and cooperating with each other, which is now one of the most widely used techniques and has many advantages. But at present there are still many problems, such as scalability, real-time recommendations and accuracy.In order to solve these problems which exists in collaborative filtering techniques, this paper adopts a cluster-based collaborative filtering technique, which clusters the highdimensional sparse user-rating matrices in advance, so as to narrow the scope of recommendations. Meanwhile, in order to improve the accuracy and stability of the clustering results, this paper conducts an in-depth study of clustering and ABC(artificial bee colony) algorithm, and proposed an improved ABC clustering algorithm with optimized initial nectar and dynamically optimized strategy.Firstly, we study the existing ideas and methods of personalized recommender system, and conduct an in-depth study of the process, modules and the adopted technology of recommender system, so as to get the technical framework of collaborative filtering technology. For the deficiency of the traditional collaborative filtering techniques, we decide to apply clustering algorithm to the technology to improve it.Secondly, taking into account the ideas, advantages and disadvantages of K-Means algorithm, combining the excellent ability of global and local optimization of ABC algorithm, we propose an improved ABC clustering algorithm. The two aspects of the traditional ABC algorithm have been improved: one is the selection of the initial nectar. Traditional ABC clustering algorithm, which has great randomness, selects K cluster centers as an initial nectar source randomly. So as to improve the quality of nectar and the efficiency of clustering, this paper propose a new way to initialize the nectar which draws lessons from K-Means++ algorithm; the other is to adjust the bees local search range. Traditional ABC clustering algorithm randomly search the area, leading to poor search capability and slow convergence. In order to overcome the drawbacks, we presents a dynamic tuning strategy, making ABC algorithm could search for different area in different evolution period. Experiment results show that the improved algorithm can improve the accuracy and stability of clustering.In order to verify the effectiveness and practicality of collaborative filtering technology based on ABC algorithm, we compare it with other collaborative filtering techniques, adopting the MovieLens datasets, respectively from the aspects of accuracy, time performance, etc. The experimental results show that the proposed method, to some extent, improves the quality and performance of recommendation.