A novel fuzzy based deep neural network for rain fall prediction using cloud images

Abstract

SummaryThe most challenging aspect of weather forecasting is predicting rainfall. In contrast, timely and precise rainfall forecasting is immensely useful in developing effective security measures for active infrastructure projects, logistics, agricultural activities, air transport operations, and flood scenarios. Artificial intelligence has been established to predict rain precipitation, but very little research has been conducted on images than time‐series data to predict rainfall. However, the accuracy of the prediction model utilizing pictures is lower when compared to time series data. Thus a novel fuzzy based deep neural network (NFDNN) model was created to increase rainfall prediction performance. Initially, our research utilized input cloud images, in which the fuzzy wavelet incorporated with neural networks is utilized to identify clouds in the sky. Second, the types of clouds are determined by a fuzzy deep neural networks cloud classifier in which the adaptive fuzzy sparse representation is used. Then, the density‐based spatial clustering for applications containing noise clustering method was used to detect rainy clouds by evaluating the cloud's color and density from the fuzzy deep neural network cloud classifier. Finally, rainfall was predicted using information such as cloud categorization, appearance, and height obtained from cloud position and cloud types. As a result, the proposed NFDNN model gives a better prediction result than the existing methods.