Research On Recommendation Algorithm And Campus Object Search Engine Development

In recent decades, the rapid development of the Internet, which not only leads to the effective dissemination and circulation of Information, but also has a profound impact on scientific research, industrial and commercial development, even everyday life of people. With the expansion of data, the Internet has entered a big data era. In such a background, the period of time for people to get information is becoming more and more longer. And a good way to access information would become a rare case. In most cases, to make a choice needs a lot of manpower and material resources, but a little success. In this paper, we introduce the importance of recommendation systems and search engines in our daily life. There are two parts in this paper which describe two different research directions.Firstly, the article introduces the development status of recommendation system. And we research the application of higher matrix factorization model in movie recommendation, a rating prediction task. With the success of Netflix Prize, application of the matrix decomposition technique in the movie recommendation becomes more popular. And after the KDD2012Cup, we are expected to know the accuracy of prediction in the movie recommendation which uses the higher matrix factorization methods. Simply, matrix factorization extracts a couple of factors from the rating matrix which called latent feature that we can use to indicate a user or an item vector. Because of non-intuitive understanding for actual meaning of the vector, it is called latent factor vector. Therefore, we define it as Latent Factor Model(LFM). After studying the basic LFM, we propose a concept called latent group factor which distinguished by the user’s interest. We call the new model ClusterLFM, and the experimental results show that ClusterLFM can improve the precision of the rating prediction.Secondly, we focus on vertical search engine, especially on the development efficiency and performance. We make,a detailed research on the development status and related technologies of vertical search engine. On this basis, we developed a vertical search engine that we. call campus object search engine(COSE), and COSE is a platform that provide search services based on campus information. What does the object mean here? We regard teachers, labs, and organization as object. With entity search technology, we can provide richer and more comprehensive information. In the research, design and implementation process of the entire framework, this article, from the perspective of engineering development, proposed a complete and efficient development specifications, with combining apache virtual host and subversion technology.