Novel Search and user interest analysis technology

Intermediary transaction http://www.aliyun.com/zixun/aggregation/6858.html ">seo diagnose Taobao guest cloud host technology Hall

One, why do we need user interest analysis

The core goal of the website is to provide a valuable service to the user and realize its value. Clearly, the needs of our users are at the heart of our concerns. In the case of the novel search provided by the bean paste net, the user's interest is the "taste" of the book that the user is interested in. For different users to search the same keyword, the search engine output the same result is difficult to meet the needs of everyone, because everyone's interest is different. Users want search results to satisfy their own "taste", which requires the search engine for the results based on the user's interest in the rearrangement.

For sales sites, good and user interest analysis can effectively improve the user's purchase volume.

Two. What applications use user interest analysis techniques

(1) Youku and other video sites: Each video page will give a recommended video list, after the video playback will also display a list of recommended videos.

(2) Taobao and Amazon shopping sites: When you buy a product, you will provide a list of items you may also buy.

(3) Dangdang and other books purchase sites: and (2) similar. When you buy a book, recommend other items.

(4) Carrefour and other large supermarket chains, according to the user's purchase record, the position of the shelves to rearrange. One textbook example is the fact that people who buy beer often buy their own diapers, according to data analysis. After analysis of specific cases, it is often found that housewives buy beer and they have a great chance of buying a diaper.

(5) Google personalized search (personalize searching): Based on past search and click Records to rearrange the results of the query. (searching result reranking), interested readers can search for scholarly literature based on the keywords provided above.

Three. The main realization technology of user interest analysis

(1) According to the user's past record and the text of the unread data matching method, will match the highest recommended to the user, such as Youku video at the end of the recommendation.

(2) Collaborative filtering technology is the use of a large number of public historical records, analysis of the relevance of two entities, and the user history of the highest correlation with the entity recommended to the user. In the case of red bean paste novel search, the system maintains a bookshelf for each user. The books in this bookshelf are a list of the books that users have been paying attention to. If a large number of users are collecting a, and the collection of book B, then we can think that books A and B style and content is basically similar, and its target for the breakdown of the readership is basically the same, we will recommend B to only search the hidden a user, a recommended to only search the hidden B users.

(3) User interest analysis based on theme model (Topic models). The problem is modeled using a subject model such as pLSA and LDA. Calculates the attribution of all books to each subject, as Vector x, and calculates the attribution of the user to all subjects, as Vector y. Calculates the Cos similarity of A and B. And will be the most similar to a certain number of books recommended to the user.

Red bean paste Novel search (http://www.docshare.org) provides, welcome to reprint, but please do not arbitrarily modify, lest distort the meaning of the original text. (My last article was reproduced and simply replaced the keyword, resulting in a full error)