| Summary: | 碩士 === 國立彰化師範大學 === 資訊工程學系 === 106 === There are several problems in traditional query of books at online bookstores. First, different books may have different degrees of satisfaction to users’ query requirements. Traditional query of books is limited in its abilities to come to grips with the issues of fuzziness. Second, different conditions may have different degrees of importance in users’ opinions. Traditional query of books cannot differentiate the importance of one condition from that of another. Third, several conditions cannot be considered simultaneously to rank a book based on their degrees of satisfaction and degrees of importance. To alleviate the mentioned problems, we propose a fuzzy query system for books through data mining and deep learning. First, a crawler is developed to gather and analyze books at online bookstores and libraries. Deep learning is applied to classify books into categories. Second, a data mining approach is proposed to obtain the degree of correlation between words in book data. We utilize mutual information to compute the degree of correlation between words, and utilize association rules to mine keywords for books. Third, a mechanism is proposed to state fuzzy queries by fuzzy conditions and to differentiate between fuzzy conditions according to their degrees of importance. Finally, the recommendation of books is made by collaborative filtering. The similarity of users and the popularity of books can be computed based on the clicking, collecting, and sharing of books by all users.
|
|---|
