Precipitation prediction by integrating Rough Set on Fuzzy Approximation Space with Deep Learning techniques.

Artificial intelligence has evolved into a mechanism that can predict environmental changes and investigate climatic conditions. In recent days, extreme precipitation and flooding have resulted in huge damage to the society of human beings and farmers. In real-life, the climatic conditions were predicted using huge collected data. However, the collected data are not utilized properly due to the presence of uncertainty. Most of the research was carried out in time series data for classification and prediction. However uncertain data leads to erroneous predictions. Thus, an integrated model is required to overcome the uncertainty and to improve the prediction accuracy. The paper aims in integrating the Rough Set on Fuzzy Approximation Space (RSFAS) with a Deep Learning (DL) techniques to predict the precipitation level in the southern coastal areas of India in a seasonal way. The proposed model can handle uncertainty using an RSFAS whereas the DL technique is utilized for classification and prediction. The developed model is estimated based on various evaluation metrics and optimizers, benchmarked with existing Deep Learning and Machine Learning (ML) techniques to demonstrate the prediction accuracies obtained with minimized error rate. Consequently, the proposed model assists in understanding the climatic conditions in extreme events to provide a prior warning about heavy rainfall to the people who live near the coastal areas. • Towards the development of a low-cost solution for seasonal precipitation prediction. • Rough Set on Fuzzy Approximation Space with LSTM used for precipitation prediction. • High accuracy as impact of tuning hyperparameters with adam optimizer for prediction. [ABSTRACT FROM AUTHOR]