With the fast increasing number of Web databases (WDBs), it is core issue in the study of Web data integration that we should select the most appropriate composition of databases to query and obtain more targeted data at a smaller cost. In this study, in order to reduce redundant data from different sources, we propose a novel method of Web databases redundancy computing to select proper Web data sources for given keywords. To solve the problem, we propose a web database feature representation model, and based on sample data from the sources, we put forward the deep web redundancy computing method considering three different data types: text attribute, numeric attribute and categorical attribute. Experiments show that this method can achieve the desired objectives and can meet the demand to the integrated system very well.