Research On Data Mining Technique Using Artificial Neural Networks

Data mining technique represents extracting a sequence of unknown, valid and operable knowledge from a great deal of datum; it is an important step in knowledge discovering procedure. The principle of discovering unknown knowledge by data mining technique is different from such data processing methods like proving validation after proposing hypothesis. The availability of mining results lies in its correctness and reasoning; the operability lies in its usage in decision supporting. Data mining technique is widely used in every scientific aspect in our society.Classifying technique is one of the utmost valuable ones in application fields. Datum classification represents extracting the together characters from a group of objects, and classifying them into different class according to the result model. This model can map a record in a certain database into one of the given class. The procedure of datum classification consists of two steps: building up the datum model and classifying datum by it. But before classifying, the correct rate should be evaluated, once acceptable, unknown data record and object can be classified into different class.Artificial neural network is a simulating brain processing net system based on modern neural biologic researching. It can not only processing ordinary numeric datum, but also has the ability of processing knowledge, learning and memorizing. Procedure of data mining based on neural networks consists of three steps: datum preparation, rules extraction and rules evaluation. This paper was discussing two kinds of rule extracting algorithms, pedagogical and decompositional algorithms. After presenting decompositional algorithm RX, computing the associate rate of input and output datum, then RBF neural network based on JOONE framework was used for selecting procedure. This procedure was to simplify the structure of the network, reduce the number of input nodes. At last, by two UCI datasets, this paper has proved the availability of this method. But there is no one classification method was the best one for all the problems, more efforts are to be spared in such fields.