Abstract:Large amounts of data on the Internet were hidden in the deep Web. Traditional topic crawler could only get surface data, and crawl the deep data was powerless. This paper established simulation model of Event-Driven, Used improved filtering algorithm based on BM by links, and content filtering algorithm based on VSM(vector space model), with model parameters of car, used Htmlparser to analysis dynamically generated pages, processed their structures to store in database, thus achieve extraction and parsing of car's information. The results of experimental show that the model for the same field datas has a good event trigger adaptability and high filtering accuracy.