The invention relates to a focusing relevancy ordering method for a
vertical search engine. Aiming at the problem that a focused crawler cannot pass through a dark tunnel, the invention improves a focused
crawling strategy of the focused crawler by using an on-line learning method and utilizing an auxiliary function, so as to lead the focused crawler to capture subject data with higher relevancy. A
PageRank algorithm and an
improved algorithm thereof are studied, the webpage clicking action of a user is modeled, and the transferring way of a
PageRank value among links is improved, so as to put forward the
improved algorithm. As to the
disadvantage that the dimensionality of a
feature extraction model of webpage weight is over high, a user-defined method of the webpage weight is put forward, so as to define a factor of the webpage weight and measure the weight of the factor of the webpage weight according to the divisibility criterion, thereby providing an
evaluation function of the webpage weight and effectively lowering the dimensionality of a webpage feature space. By utilizing the method in the invention, the user can obtain a high-quality search
result set when using a subject resource
search engine system.