Research On Deep Web Data Merging Based On Neural Network

With the rapid development of the World Wide Web, the amount of useful information on the Internet has ever been increasing. In order to make effective of the information, the data from different websites need to be recognized and merged under a uniform pattern. However, the different qualities and self-determination of Web data bases make it a very challenging issue.This paper made a comprehensive review and analysis of the literature at home and abroad on deep web data merging techiniques, and proposed a deep web data merging method based on neural network according to its self-organizing, self-learning and self-adjusting properties. The main research points of the study are like the following:(1) It investigated the research situation on deep wet data merging techniques both at home and abroad, and studied carefully the several typical ones.(2) Considering the defects of previous web data merging techniques, the present study proposed a method based on neural network. This method realizes the web data merging from the aspects of data table features and property features, providing a new approach towards web data merging.(3) Data converting is an important prerequisite for the neural network based deep web data merging. Therefore, this research also studied the web data converting techniques, which map the web data onto structured relational data bases for later analysis.(4) Neural network toolbox was applied for network training and simulating, which ensures the maneuverability of this method. Several real cases were used to validate this data merging approach.Finally, the research work involved in the thesis was summarized, and several suggestions for further studies on deep web data merging were also proposed at the end of the thesis.