Study Of Data Mining Method Based On Neural Network

Data mining is a kind of new technique that discovers and extracts the information hidden in large database or data warehouse, it can make automatic analysis and inductive inference to the data in data warehouse, and find the latent pattern; or create association and establish new business model to help the leader make correct decisions. Data mining research and practice indicate that knowledge is hidden in the innumerable data accumulated day by day, yet it can't be discovered only by complex algorithm and inference. The technique of artificial intelligence, especially the combination of neural network and data mining points out a new way for the research of data mining theories and methods.This paper surveys the present situation, development tendency and application domains of data mining and neural network, then expounds the basic concepts, key models and traditional realization methods of data mining and neural network. Aim at the concrete study content of data mining, this paper brings forward and carries out the general processing methods of data preparation in data mining using neural network, the construction of classification decision tree based on neural network, classification and prediction based on neural network, incidence rule mining based on neural network and the models and algorithms of cluster analysis based on self-organization neural network. The methods of data mining based on neural network make advances in information processing capability, incidence rule mining and representation capability comparing with traditional data mining methods.In light of the combination of research outcome of data mining based on neural network and the scientific research project of the third oil recovery factory in Daqing oil fields, we have developed the system of water flooding layer identification of grape flower oil-bearing sand in the north of SAERTU development area. The bottom data support of this software system is oilfield prove development database, and the software mainly consists of data cleaning, data conversion, data extraction, data mining based on neural network, result assessment and knowledge representation, realizing the automatic identification of water flooding layer.