Support Vector Machine Algorithm For Data Mining Research

Data Mining is a new kind of scientific tools used for knowledge discovery in database. It becomes a hot topic in relative domains during past few years. Intelligent machine learning algorithms play an important role in DM systems. Researches on outstanding algorithms are indispensable to DM works, with advantages to performance and efficiency.Support Vector Machine is a tool based on statistical learning theory, widely used for prediction tasks on small sample data for its generalization capacity. It has been used in various domains since its birth. To implement applications of SVM on DM will be a sensible topic of research.In this thesis, SVMs are used as tools for DM tasks.The main works are as follows:1. A review of models, tasks and methods in DM, with SVM related theories, algorithms and techniques is presented.2. Tasks of classification and regression are solved by SVM based tools. Data used in experiments come from small experimental samples and large machine learning databases. Problems of performance, both models and samples, are discussed subject to experiment results.3. Wavelet functions are used as kernel functions of SVM, called Wavelet SVM. Analysis on SVM kernel functions is presented.4. A Proposal of SVM parameters optimization is presented by genetic algorithm, with experimental proofs on its reliability.Deductions and Experiments in this thesis are based on standard methods or tools, and most of them are repeatable. The contents of this thesis can be helpful to relative researches or applications.