Margin Maximization Model of Text Classification Based on Support Vector Machines

Support vector machines (SVMs) are more suitable for text categorization than traditional machine learning methods by acknowledging various statistical characteristic of text learning task. By introducing the margin maximization principle in the statistical machine learning theory, the feature statistic matrix based on average document mapping (FSM-ADM) model, which partitions the set of features using weighted odds ratio, is proposed in the form of generalization capability estimation theorem with rigorous proofs and solid experimental validation. The theoretical model has successfully discovered the unexplored capability of being classified in text classification