Based On Self-organizing Map Networks And Perceptron Data Mining Methods And Applications

With the widely application of database and the sharp development of Internet, it is an urgent problem to mine useful information from the large database or data warehouses. Therefore, DM (Data Mining) technology is developed rapidly to meet the need. But DM often faces so much data which is noisy, disorder and nonlinear. Fortunately, ANN (Artificial Neural Network) is suitable to solve the before-mentioned problem of DM because ANN has such merits as good robustness, adaptability, parallel-disposal, distributing-memory and high tolerating-error.This thesis briefly expounds the basic concepts of DM and ANN, including concept, preprocessing and algorithms of DM and the basic neuron model, topology and the learning of ANN. Data preprocessing, including data cleaning, data integration, data transformation and data reduction, is also discussed.Clustering is the task of grouping the objects of a database into meaningful subclasses (that is, clusters) so that the members of a cluster are as similar as possible whereas the members of different clusters differ as much as possible from each other. Due to its unsupervised learning ability, clustering has been widely used in numerous applications, such as pattern recognition, image processing, market research and so on. Clustering can find out the dense or sparse areas of data distribution, which can help to discover the distribution mode and interesting relationship from data.The main clustering algorithms exiting are analysed in the thesis. Then self-organizing map network is researched and the characteristics of the network are discussed. We optimized the self-organizing map algorithm, improved the speed of clustering phrase. After that, the GHSOM network is discussed. We proposed a new GRAGHSOM algorithm by using grey relational analysis. Experiment results showed that the GRAGHSOM algorithm has better performance in high-dimension data clustering.At the end of the thesis, perceptron network is discussed, including single layer perceptron and multilayer perceptron. Today, Local weather forecast(LWF) are very important in flight training. So, the perceptron network is applied to LWF for flight training.