Improving River Runoff Forecasting through Anomaly Detection and Repair

Accurate river runoff forecasting plays an important role in management and utilization of water resources. However, most of time series forecasting methods for runoff data are sensitive to anomalies or outliers. Anomalies incur negative effects on the accuracy of time series prediction. Recently, there have been some research works which proposed a novel framework to improve time series prediction through anomaly detection and repair in a preprocessing step before forecasting. In this work, we try to apply this approach to improve the predictive accuracy of river run-off forecasting. Experimental results on four run-off datasets from three rivers in Vietnam reveal that the proposed approach can improve the accuracy of run-off prediction with two forecasting methods: k-nearest-neighbors and ANN.