Harnessing machine learning for streamflow prediction: A comparative study of advanced models in the Upper Klang River Basin, Malaysia

Study region: Upper Klang River, Wilayah Persekutuan Kuala Lumpur, Malaysia. Study focus: This research focuses on assessing the effectiveness of various machine learning (ML) techniques for predicting daily streamflow in the upper Klang River, Malaysia. The study emphasizes model development, data preprocessing (such as imputing missing data with Kalman smoothing and selecting lagged inputs through autocorrelation analysis), and performance evaluation using statistical indicators. By identifying the exponential GPR as the top-performing model, the research aims to enhance the accuracy of streamflow predictions, contributing to better water resource management, disaster mitigation, and hydrological decision-making. New hydrological insights for the region: This study provides valuable new insights into the streamflow pattern of the upper Klang River, Malaysia, by leveraging advanced machine learning (ML) techniques for accurate daily streamflow prediction. The key findings emphasize the proficiency of Gaussian process regression (GPR) in predicting complex streamflow patterns. This model's ability to capture intricate temporal variations in streamflow offers promising potential for improving flood risk assessment. Additionally, the incorporation of lagged inputs, determined through autocorrelation analysis, significantly enhances the model's accuracy, underlining the importance of considering temporal dependencies in hydrological forecasting. These insights can help hydrological authorities and decision-makers refine predictive models and optimize flood mitigation strategies, ultimately contributing to better environmental and community resilience in the region.