A Method for Topic Classification of Web Pages Using LDA-SVM Model

摘要

Using the model we propose in this paper, we can use little signed samples for web pages classification. This will greatly reduce the manual work. Through the analyze of experiments, we find the connection between topic number K and data set. Further testing is needed for finding accurate correlation. Another, we get 90 thousand poor classification data set, some of the classification is wrong and some web page is outdated or change. We want to build a flexible and effective system for unknown web pages classification. This work is focus on sophisticated category, and our next work will expand the data set and categories. Wrong data filtering and hierarchical classification is both challenge.