Performance Evaluation Of Various Machine Learning Algorithms For Class Imbalanced Landslides Identification

Identification of landslides is essential for risk analysis and control. However, because landslides and land displacement are extremely rare occurrences, landslide identification faces a class imbalance issue. In this research, we try to use feature engineering to enhance model performance and apply strategies to fix the imbalance problem in the landslide identification dataset. More specifically, we applied synthetic and random oversampling strategies to the landslide dataset in order to solve the class-imbalance problem. We have applied various machine learning techniques to landslide identification dataset for identifying landslide and have evaluated these algorithms. Results showed that the random forest approach outperformed alternative classification methods. The random Oversampling Technique outperformed among class-imbalance approaches as well. Feature engineering has also helped in improving performance of most models.