Enhancing flood forecasting accuracy through improved SVM and ANFIS techniques

Extreme rainfall in upstream watersheds often results in the rise of river water levels, leading to severe flood disasters in the downstream catchment. Therefore, monitoring river water levels and flow is crucial for flood forecasting in early warning systems and disaster risk reduction. However, forecasting river water levels remains a challenging task that cannot be easily captured with classical time-series approaches. This paper explores the potential of improving flood forecasting accuracy by combining two forecasting techniques: Support Vector Machine (SVM) and Adaptive Neuro-Fuzzy Inference System (ANFIS) by simple averaging and weighted averaging methods and optimizing their contributions. To tune different individuals' weights the genetic algorithm and K-nearest neighbors' algorithm (K-NN) were used to find the optimal weight combination. The committee machine model was employed to forecast the river water level in different lead times from 1 hour to 6 hours applied to the Selangor River. Model performance was evaluated and analyzed using various performance metrics, including mean percentage error (MPE), root mean square error (RMSE), mean absolute error (MAE), and correlation coefficient (R). The results show that the proposed Intelligent Committee Machine Learning (ICML) outperformed SVM and ANFIS for most performance indicators. This method aims to develop a robust and accurate time series forecasting model by combining multiple forecasting techniques and optimizing their contributions, ultimately leading to improved prediction performance.