Enhancing Riverine Flood Forecasting capability of Gaussian Process Classifier using Hyper-Parameter Optimization

Climate change has increased the rate of melting glaciers and snow cover. The precipitation patterns have also changed along with the increased rate of melting snow and glaciers which contribute heavily to the floods in rivers. The floods are most damaging, whether we talk about the economy or human lives. This makes the forecasting or prediction of floods important. Several methods and techniques were tried for decades to predict floods. Machine Learning (ML) is the latest and most advanced technique used for forecasting and prediction in almost every field. So, in this study, we developed a machine learning model, Gaussian Process Classifier (GPC) with hyperparameter tuning by Random SearchCV to predict the flood risk in the Himalayan river Jhelum. We pre-processed our dataset using techniques like feature scaling and data balancing. Feature importance evaluation was also done using an ensemble machine learning algorithm, extra trees classifier (ETC). Our model showed a ROC/AUC score of 0.803, average precision of 0.79 and the most important metric, the recall value of 0.85.