A Deep learning-based rainfall prediction for flood management

Accurate prediction of rainfall remained a center of attraction for researchers and industries. Its importance remained significant because of its effectiveness, as heavy and irregular rainfall can have many impacts like flash floods, the destruction of crops and farms, and property damages. The hydrological model and invasion model are the two popular models used for predicting rainfall. These models make predictions from the water level through water gauges in the river and identifies the hard-hit areas from flooding. However, these models are unable to find the hidden patterns in the rainfall data. Artificial intelligence has the potential to apply computational methods by processing hidden knowledge acquired from different patterns of past weather data. Such rainfall prediction can help to make early decisions to prevent flood situations. This paper uses Long Short-Term Memory (LSTM) based Recurrent Neural Network (RNN) to predict rainfall by analyzing correlations between different attributes. Results have outperformed the other algorithms in terms of accuracy and speed.