Research And Application Of Collaborative Filtering Recommendation Algorithm

As the 21st century is the age of the Internet, the network has become one of the most important sources of information. However, with the development of information technology and Intrnet, people gradually from the lack of information era into the information overload era. At this time, the recommendation system came into being, and recommendation algorithm as its core has been widely studied by a lot of scholars at home and abroad.Collaborative filtering recommnendation algorithm as one of the most successful recommended algorithms in the field of practical application, has some advantages. For example the higher degree of the personalized recommendation, helping users find potential interested in the project, filtering content information which are more abstract complex. But as a single pattern of static algorithms, collaborative filtering recommendation algorithm inevitably has some functional defects. Despite an abundance of domestic and foreign scholars have made many achievements, but there are still some difficult problems to be solved, including data scalability issues, data sparseness problems, cold start problems, etc.Therefore, such as hybrid recommendation algorithm, several recommended mode of parallel computing technology has become current popular research direction in the field of recommendation algorithm. These recommended techniques in accordance with the special requirements of different applications, select the corresponding recommendation algorithms to take advantage of different algorithms to play the effect of recommendation 1 plus 1 greater than 2.This paper around the subject of the traditional collaborative filtering recommendation algorithm study, mainly focus on solve the problem of data extension of traditional algorithm and data sparseness problem in order to further improve the accuracy of recommendation results. For these two problems, this paper gives the improved algorithm which combined the dynamic learning algorithm model with collaborative filtering model.The main contents of this thesis are as follows:1. In view of the traditional collaborative filtering recommendation algorithm based on the project of data sparseness problem and recommend the low accuracy problem, the thesis puts forward Hidden Markov Model and the traditional collaborative filtering recommendation algorithm based on the project of mixture of "hybrid recommendation algorithm based on the HMM ". The algorithm using Hidden Markov Model to all the asers in the system evaluation behavior and the history of the target user behavior to carry on the overall analysis, tofind the probability of the next moment a group of users with the highest score object, and the probability of occurrence of these scores with traditional objects project weighted similarity calculation method to get a new recommendation similarity ultimately produce results.2. To solve the problem of high computational cost and the low recommendation performance when user preferences change over time in traditional user-based collaborative filtering recommendation algorithm, we propose an algorithm with the combination of hidden Markov model and User-based collaborative filtering recommendation algorithm to correctly interpret the users’ product selection behaviors and make personalized recommendations. In this algorithm, the user preference is modeled as a hidden Markov sequence, so we can filter and find the neighbor which has a high matching score with target user. In this way, we reduced the computational cost. And with the combination of hidden Markov model and User-based collaborative filtering recommendation algorithm, system can avoid the low performance of the recommendation and make personalized recommendations when user preferences change over time.