Analysis And Implementation Of Video Recommendation System Based On Web Mining

With the rapid development of the Web2.0 technology, it’s very difficult for Internet users to simply rely on database searching and the statistical analysis to obtain the information that they need win the face of the network’s big data, and the information overload lowers users’ efficiency of using information. Therefore, it is imperative to change data, which are for futher process, into information with potential values in an automatic and intelligent way, so as to serve users’ demands.This paper analyzes and designs a video-recommending system that is based on Web mining. First of all, it analyzes user’s web logs that have been gathered to obtain the user’s behavior and attribute data of watching videos. This data includes user’s registration information, records of videos searched by him or her, records of watched videos and the user’s scores for videos. Later, it discovers the user’s preferences according to the matrix relationships between the records of the user’s selecting, searching, watching and scores of the video projects, namely, establishing the user’s interest model, so as to find user’s preference. In the model, the similarity calculation formula in the collaborative filtering algorithm is modified, so as to proactively recommend to the user videos that meet his or her interests. The specific work and innovations in this paper include the following contents:Firstly, based on existing research results, the current individual recommending system has been analyzed, compared and concluded. Meanwhile, several different individual recommending systems that have the most widespread application at present are introduced, and some improvement to the original individual video-recommending system by combining the Web mining technology is proposed.Secondly, the Web mining technology has been introduced into the traditional video-recommending system, at the same time, the user’s explicit data and implicit data is obtained, which includes the records of the user’s scores for videos and records of videos searched by the user. The categorical regression is used to establish the model of user’s preferences. Moreover, this experimental system adopts the similarity calculation formula and applies it to punish the influence exerted by popular videos in the system on the recommendation.Then, targeting at the systematic characteristics of the demands for individual video-recommendation, this paper carries out the design from the system’s target users and systematic demands & functions, including the detailed architectural design, the data-processing platform’s architectural design and the database design, providing a clear thinking for the process of developing the individual video recommending system.Finally, the video-recommending system that is based on Web mining is achieved and tested. The experiment shows that: discovering user’s preference information based on Web mining is objective. It’s more accurate and innovative than simply relying on user’s score data, and is able to alleviate the cold-star problem in the recommending system which results from new user or no score for new videos, so as to improve the quality of users’ experience of watching videos.