Predicting Flood Streamflow with Auto Regressive Integrated Moving Average Models

Flooding is the most common natural disaster and continues to increase in frequency intensity due climate changes [7]. Currently, there a lack of efficient tools predict flooding. This research aimed create Time Series Machine Learning (ML) program using Auto Regressive Moving Average (ARIMA) models forecast streamflow, one prominent factors flood prediction. A streamflow dataset from Ganges River, Bangladesh was used plot several graphs river Log Volume observe possible trends. Another graphed check quantify how much distribution stream volume changed over course 10 years KL Divergence. The analyses Partial Autocorrelation Function (PACF) (ACF) tests were help obtain ARIMA parameters (p, d, q) as (1, 1, 1). However, forecasted function not accurate when compared with previously recorded data because heavy seasonality. As result, final redesigned Seasonal (SARIMA) account for inaccuracy. SARIMA model subsequent close actual data. Such accuracy indicates that this method can be useful tool navigating preparing floods.