An Orthogonal Selection Method for Identification Features of Relative Poverty Based on Pairwise Sample Comparison

To solve identification feature selection problem of relative poverty, an orthogonal selection method based on pairwise sample comparison was proposed. Two types of paired sample sets were collected by means of pairwise comparison, one was the paired samples of “existing relative poverty” and the other is the paired samples of “non-existing relative poverty”. Then, a new feature subset evaluation function was constructed based on the idea of pulling similar samples closer and pushing dissimilar samples further apart. Finally, a relatively simple orthogonal experiment was employed to select features. To validate the effectiveness of the method, 356 registered poor households and 212 non-registered poor households from the Dabie Mountain area were considered as research subjects. Four sets of paired sample sets were randomly constructed to select out four sets of key features. Various classifiers including logistic regression, decision tree, support vector machine, deep neural network, random forest, boosting, and naive Bayes were used to evaluate the identification performance of these key features. The results indicate that Six classifiers, logistic regression, support vector machine, deep neural network, random forest, boosting and naive Bayes, adopt four groups of key features for identification, and the identification accuracy, sensitivity, specificity and AUC value can all exceed 90%. There is little difference in the identification performance of these key features selected by different paired sample combinations. The identification performance of all four groups of key features can reach the identification performance of all features. The proposed method is simple in principle and easy to operate, and can select out the identification features of relative poverty in the absence of relative poverty classification standards or difficult to formulate the classification standards.