Machine Learning And Gis Techniques In Flood Risk Mapping-

Case Stady Of M’sila

Abstract

 Geographic Information Systems (GIS) play a vital role in providing and designing flood maps and their probability, especially through their integration with modern machine learning (ML) techniques. The aim of this research paper is to highlight the importance of the integration between these two technologies in drawing flood risk maps. In order to achieve this goal, we studied the case of the city of M'sila, which is one of the areas in Algeria most prone to this type of hazard. The study resulted in drawing a map using GIS that identifies the most flood-prone areas in the city of M'sila and achieving the desired goal, after comparing the results of three machine learning algorithms, which are the extreme gradient boosting (xGboost) model, the random forest algorithm, and the k-nearest neighbor algorithm. Each of these algorithms gave a certain level of accuracy, with the latter reaching 97.92%, 95%, and 93.75%, respectively. Finally, the first model, which gave the highest accuracy, was used in the process of drawing the final map, which is the xGboost model.