Anomaly Detection Using a Sliding Window Technique and Data Imputation with Machine Learning for Hydrological Time Series

Water level data obtained from telemetry stations typically contains large number of outliers. Anomaly detection and a imputation are necessary steps in monitoring system. can be detected if its values lie outside normal pattern distribution. We developed median-based statistical outlier approach using sliding window technique. In order to fill anomalies, various interpolation techniques were considered. Our proposed framework exhibited promising results after evaluating with F1-score root mean square error (RMSE) based on our artificially induced points. The present system also easily applied patterns hydrological time series diverse choices internal methods fine-tuned parameters. Specifically, the Spline method yielded superior performance non-cyclical while long short-term memory (LSTM) outperformed other distinct tidal pattern.