Classification Algorithm Based On Logistic Regression And Support Vector Machine

Classification is one of the basic topics in data analysis and machine learning domain. Lots of results have been achieved in this field. Logistic regression model is a widely used technology of multiple quantitative analysis with good robustness and model explicable. Support vector machine, based on statistical learning theory, has shown some special advantages in small sample, nonlinear and high-dimensional problems. Therefore,SVM has become a hotspot in the field of machine learning recently.Traditional Logistic regression methods commonly take 0.5 as demarcation point, which might cause great classification error risk, especially the sample points of fuzzy intervals near 0.5. In order to improve logistic regression, an integrated algorithm with SVM is proposed in the thesis. This algorithm reduces the classification error of logistic regression, applying the output of SVM.In this paper, it first outlines logistic regression, the principle, derivation and model checking. Then, SVM and related theory are introduced in detail, including the basic theory of machine learning, statistical learning theory, SVM classification algorithm and model parameter selection.Based on the theoretical analysis of logistic regression and SVM, an integrated binary-class classification algorithm is proposed. The validity of this integrated algorithm is illustrated by numerical results on several datasets.At last, we summarize the research of this paper and put forward some suggestions about further study.