Research On A Prefetch Policy For Deep Web Data Integration System

With the graduate maturity of Web techniques and the fast increase of Deep Web's containing data, Web database is turning into the main way for people to obtain information. As a result, the study on Deep Web is attracting more and more attention today. This paper aims at posting a prefetch strategy based on Deep Web integrated environment. With prefetch technique applied into the accessing course of Deep Web data, users would get not only higher quality information but also faster response since the bad influence of network delay could be reduced.Prefetch technique is a kind of proactive Cache technique, the cached information contains not only the visited information but also those unvisited. With prefetch technique, the data which users are going to visit would be extracted in advance and stored into the web cache when users are visiting the present content. Since the transmission of prefetch data takes the spare time of system, it could be carried out when users are reading and thinking. The improvement which prefetch technique brings about is reducing the delay time users could feel. The existence of time and place limitation for user's visiting provides direct basis for the study of prefetch technique.Meaning caching technique is one of the optimization methods for data base, making up for the deficiency that tuple caching and page caching can not sufficiently support relational database. As a result, it is of plenty study value. With Meaning caching technique, the searching result and relative meaning would be cached, which would provide answers to the future search. This technique reuses cached data based on the meaning locality, namely the correlation of searched resultsWith the combination of prefetcthing technique and semantic caching, this paper posts a prefetching strategy which is appropriate for Deep Web data integration system. According to the access features of Web database, firstly, properly define semantic cached items and their matching types which based on the Deep Web; then, corresponding semantic cached items are generated for queries putting forward by users. At the same time the access frequency of each semantic item is calculated according to different required matching types and through polynomial regression model, the visit possibility of each semantic item's next period is predicted. Those semantic items to be prefetched are obtained under the forming condition of prefetching queue.Therefore, on every new visit cycle, users'interested data will be stored in the cache in advance for further searching use. Experimental results show that adding prefetching mode to deep web data integration system will largely reduce the inquiry respond time of users, meanwhile, the prefetching accuracy and network flow will be improved along with the increase of caches. However, when the number of caches reach to a certain extent, the prefetching accuracy will decline to the opposite, therefore, when selecting the right number of catches, we should comprehensively take every aspects into consideration to avoid influencing the performance of network causing by network traffic.