Rules Extraction From Artificial Neural Networks For Classification Based Improvedant Colony Algorithm

As the development of information era, the information which people get is growing very fast, because of this, people are eager to find more useful and correct knowledge from the giant data, and that makes the data mining be proposed. Data mining is to find the connects the data owner not notice and conclude the data with more understandable and novel ways by finding methods to analysis the data sets ,it obtained large application in such as economic and scientific fields. Data mining can help users only need to pay attention to focusing most important information.Classification obtains great concern in the field of data mining. Its main purpose is to predict the classification of data objects. Because of the excellent performance that artificial neural network (ANN) can obtain from prediction, studying from experience and generalizing from the previous samples, making it an important method of classification. Although ANNs can achieve high classification accuracy, its explanation capability is very limited, as to restrict its application. Therefore one of the main challenges in using ANNs in data mining applications is to extract explicit knowledge from them.So a novel approach is proposed in this paper for generating classification rules from ANNs. ANN rules extraction techniques are divided into two major categories of decomposition method and pedagogical method. This paper is based pedagogical method. Although there are several approaches proposed for classification rules extraction from ANNs, this present approach is fundamentally different from them. This paper presents an improved ant colony algorithm based for classification rule extraction from ANNs.An improved ant colony algorithm is to help extracting rules, and binary way is used to deal with continuous variables, also improved ant colony algorithm can solve the ANN's limited explanation capability to extract rules from the data. This method can be used in all kinds of ANN based classifier, but do not depending on structure of nets and the training algorithm.Experiments show that this approach has a big potential and could coordinate neural network to obtain more accurate and concise rules of classified data well.