SVM initiative learning algorithm research for meta-information obtaining

The crawler technique adopts as same obtaining technique as common URL instead of telling the difference between webpage URL and seed file URL when the traditional crawler technique crawling the seed file, so it cannot meet the requirement of active monitoring model both in efficiency and accuracy. The traditional SVM classification algorithm was improved as per the requirement of active monitoring model: classified URL by powered least squares support vector machine, and directly obtained the seed file URL by file-download technique as per the classified result. The contents of improved classification algorithm are summarized as follows: collecting the sample information, discretizing and normalizing its feature vector and obtaining it; then obtaining the key parameter of the algorithm by learning it; finally classifying URL. As the test indicates, this improved algorithm can meet the requirement of active monitoring model.