The Cumulative Logistic Regression Classification Of Students' Poverty Data

Nowadays, poverty ranks for students from China's universities is mainly based on subjective judgments. In order to explore a more reasonable way to define family economical levels of students , this paper applies cumulative logistic regression method to in questionnaire poverty data collected in our university, and using Bagging method to improve the degree of accuracy of classification. 10 times cross-validation results shows that: the accumulated logistic regression data classification accuracy for poverty data is 55.224%; the model distinguishes the category 2(poorer) and the category 3 (generally poor) not so obviously, whereas makes a relatively good distinction between the category 4(not poor) and the category 1(especially poor); misclassifications mainly occur in the adjacent categories mistranslations. By cumulative logistic regression method in using bagging algorithm in the data, the accuracy improves little. Finally, this paper also analyzes the reason why the low classification accuracy of the data in the aid of office workers and other experts teacher's experience and advice.