Duplication Deletion Mechanism of Web Search Engine Based on Multi View Canonical Correlation Analysis

摘要

Spam is the important problem that may be faced in information retrieval field, which seriously affects the user to obtain relevant information. First to divide spam website features into two different views, views based on the content features and views based on link features, by using canonical correlation analysis and the related improved methods for feature extraction to generate two sets of new features. Then it adopts different combinations on the two-view new features to produce single view data, and uses this set of data as the training data to construct the classification algorithm. The experimental results show that the spam web can be regarded as two-view data, and applied in multiple-view canonical correlation analysis technology, which can effectively improve the identification precision of the spam web.