Automated rain fall prediction enabled by optimized convolutional neural network‐based feature formation with adaptive long short‐term memory framework

Abstract The main concept of this article is to plan for the intelligent rainfall prediction using the combination of deep learning models. The dataset is gathered from the standard publically available dataset concerning the Tamil Nadu state. The collected data is given to the feature extraction, in which few features such as; “minimum value, maximum value, mean, median, standard deviation, kurtosis, entropy, skewness, variance, and zero cross” are extracted. Additionally, the extracted features are applied to the optimal feature formation, in which optimized convolutional neural network (O‐CNN) is employed for the final feature formation. Here, the activation function, count of pooling layer, and count of hidden neurons are tuned with the intention of minimizing the correlation between the selected features. Once the optimal features are selected with less correlation, adaptive long short‐term memory (A‐LSTM) is adopted for the prediction model. Here, the enhancement is concentrated on minimizing the function concerning the error through the optimization of the hidden neurons of A‐LSTM. The improvement of both the deep learning models O‐CNN and A‐LSTM is performed by the improved sun flower optimization (I‐SFO). The research results reveal superior performance to existing techniques that offer novel thinking in rainfall prediction area with optimal rate of prediction.