AcoNeural: A Hybrid Neural Network Model for Predicting Daily Rainfall

Daily rainfall prediction is essential for a broad variety of applications, consisting of water resource handling, agriculture, and getting ready for disasters. Although neural network models have been proven to be successful for forecasting precipitation, successful implementation requires careful choice of model architecture and training parameters. The central focus of this article is to make and contrast several Neural Network models, including Narrow Neural Network (NN), Wide Neural Network (WNN), Medium Neural Network (MNN), Bilayered Neural Network (BNN), and Trilayered Neural Network (TNN) with Ant Colony Optimization (ACO-NN) for predicting daily rainfall in Alipurduar, West Bengal. For this, many essential weather factors, like temp, wind speed, humidity, and many more, are considered inputs, and day-to-day precipitation was used as an output. Evolution Metrics like mean square error (MSE), mean absolute error (MAE), root mean square error (RMSE), and R2 are used to track the algorithm’s efficiency. The results indicate that all models did an excellent job of making predictions, but ACO-NN did the best compared with the others in terms of RMSE. This implies that the ACO-NN methodology has potential as a viable solution for the prediction of rainfall.