A novel fuzzy K -nearest neighbor inference model with differential evolution for spatial prediction of rainfall-induced shallow landslides in a tropical hilly area using GIS

摘要: This research represents a novel soft computing approach that combines the fuzzy k-nearest neighbor algorithm (fuzzy k-NN) and differential evolution (DE) optimization for spatial prediction of rainfall-induced shallow landslides at tropical hilly area Quy Hop, Vietnam. According to current literature, k-NN DE are state-of-the-art techniques in data mining have not been used landslide. First, database was constructed, including 129 landslide locations 12 influencing factors, i.e., slope, slope length, aspect, curvature, valley depth, stream power index (SPI), sediment transport (STI), topographic ruggedness (TRI), wetness (TWI), Normalized Difference Vegetation Index (NDVI), lithology, soil type. Second, 70 % were randomly generated building model whereas remaining 30 % validating model. Third, construct model, search optimal values strength (fs) number nearest neighbors (k) two required parameters k-NN. Then, training process performed obtain Value membership degree class each pixel extracted be as susceptibility index. Finally, performance capability assessed using classification accuracy, under ROC curve (AUC), kappa statistics, other evaluation metrics. The result shows has high dataset (AUC = 0.944) validation (AUC = 0.841). compared with those obtained from benchmark methods, support vector machines J48 decision trees. Overall, performs better than trees models. Therefore, we conclude is promising tool should mapping landslide-prone areas.