APPLICATION OF DIFFERENT MACHINE LEARNING ALGORITHMS FOR FLOOD PREDICTION ON THE RIVER NIGER COASTS: A CASE STUDY IN NIAMEY, NIGER

Flooding is one of the most frequent and destructive natural hazards in the world, especially in urban areas. The city of Niamey, Niger, being no exception, suffers from these floods almost every year, and the risk areas are still not determined with the adequate precision that should be. It is within this framework, and to remedy this problem that we have decided to conduct work aimed at predicting the risk of flooding and determining the areas at risk in the city of Niamey, capital of Niger. To carry out this work, we used GIS and Remote Sensing tools, as well as a set of machine learning algorithms, namely K-Nearest Neighbor (KNN), Random Forest (RF), Support Vector Machine (SVM), Decision Tree (DT), and Logistic Regression (LR). Several data were used for the realization of these processes namely elevation, aspect, slope, curvature, stream power index, sediment transport index, topographic roughness index, topographic wetness index, distance from the river, geology, soil, land use/ land cover, and rainfall were selected. All of these data were pre-processed in ArcGIS and ENVI and then exported to apply Machine Learning algorithms on 2000 points (1000 flood points and 1000 non-flood points) selected randomly. It turns out that after applying the different algorithms, the RF gives the best accuracy with 93%. The results of this research show that this approach is very suitable for flood risk management in this specific area, and could be extended to other horizons.