Study of Landslide Susceptibility Mapping Based on Fractal Dimension Integrating Support Vector Machine with Index of Entropy Model

Landslides occur frequently on the Loess Plateau in the north of Baoji City, Shaanxi Province, which seriously threaten the economic development, production and life of the local people. Based on fractal dimension, entropy weight model (IOE), support vector machine model (SVM) and two hybrid models, namely F-IOE and F-SVM, are used to quantitatively predict the possible occurrence range of landslide. First of all, 179 landslide samples were used to make landslide cataloguing maps, 70% (125) of the landslide samples were used for training, and the remaining 30% (54) were used for testing. Then, 12 kinds of landslide influence factors are extracted, information gain rate and fractal dimension of each factor are calculated respectively, and four landslide vulnerability zoning models are established using training data. Finally, the performance of the model was tested using the receiver operating characteristic curve (ROC) and statistical indicators including positive predictive rate (PPR), negative predictive rate (NPR) and accuracy rate (ACC), and the generalization of the model was compared. The results show that F-SVM model has the highest PPR, NPR, ACC and AUC values in training and test data sets respectively, followed by F-IOE model. Finally, F-SVM model is the best among all models. Therefore, the hybrid model based on fractal dimension has more advantages than the original model, which can provide reference for local landslide control decisions.