Application of the borderline-SMOTE method in susceptibility assessments of debris flows in Pinggu District, Beijing, China

Abstract

According to information provided by the residents of Pinggu District, Beijing, 79 gullies were investigated. Based on the results of field investigations, 65 gullies were classified as debris flow gullies, and 14 were classified as non-debris flow gullies. By using ArcGIS software, the area and the quantity information of various factors affecting the occurrence of debris flows were obtained. The frequency ratio method was used to obtain the combinations of factors conducive to the debris flow occurrences in the study area. The susceptibility of debris flows in Pinggu District was evaluated by the information value method, certainty factor method, and logistic regression (LR) method. When using the LR method, the number of positive and negative samples should be balanced; therefore, the borderline-SMOTE method was used to generate samples belonging to the minority class. By comparing the results obtained by the three methods, it was found that the results obtained by the LR method are best. This result indicates that for disaster susceptibility assessments, the borderline-SMOTE method can be used to synthesize samples. The susceptibility evaluation results show that debris flows are more likely to occur in the southeast part of Pinggu District.