Realization of big Deep Web database optimized access with low sample deviation was researched. With the sharp increasing of the storage of Deep Web database geometrically, the problem of optimized access to the database was becoming difficult. On the basis of correlation feature rule extraction, an improved optimized access method for the Deep Web database was proposed based on Graph Model Sample process and correlation rule. The access was realized by computer simulation. The access process was started with the arbitrary effective result as the access starting line. The access result was returned, and the records were obtained in the returned web page. The next query and access was implemented with the local sample database for the nest access start line. The method avoided the effects of the query interface properties. And the Graph Model Sample method could overcome the limitation of the query interface properties. Also the correlation feature extraction algorithm was researched, and it was used as the access rule in the database for the realization of optimized and efficient query. The simulation result shows that the sample deviation is stable with the value about 10% for the simulated Deep Web database. The relevant sample deviation value is trended to convergence and is near to the real value with the increasing of the access number. According to the realistic Web database, the sample bias estimation mean value is about 21% which is higher than the simulated database. Two typical Deep Web databases are taken as the researching objects, with the new method, and the access bias is lower than another Web database which without using the proposed method, result shows nice performance of the proposed optimized database access method in application.
Stratification-Based Outlier Detection over the Deep Web