论文部分内容阅读
针对主题爬行技术中的单一分类算法在面对多主题Web抓取和分类需求时泛化能力不强的局限,设计一种利用多种强分类算法形成的分类器组合,主题爬行器根据当前主题任务在线评估并为分类器排名,从中选择最优分类器分类的策略,并开展在多个主题抓取任务下的分类实验,比较每种分类算法的准确率和组合后的平均分类准确率以及对分类效率等评价指标的综合分析,结果证明该策略对领域局域性有所克服,普适性较强。
Aiming at the single classification algorithm of thematic crawling technology in the face of the limitation that the generalization ability is not strong when the multi-subject Web crawling and classification requirements are not met, a classifier combination using a variety of strong classification algorithms is designed. According to the current theme The tasks are evaluated online and the rankings of the classifiers are selected. The strategy of the optimal classifier classification is selected, and the classification experiments under a plurality of topic crawling tasks are carried out. The accuracy of each classification algorithm and the average classification accuracy after combination are compared. A comprehensive analysis of the evaluation indexes such as classification efficiency shows that this strategy has overcome the local area of the field and is more universally applicable.