Comparative Evaluation for Humidity Forecasting Using Deep Learning

Abstract

Climate change has been well-known in recent years, and it is predicted to continue in the future. Generally, predicting weather has a great effect on citizen’s private life in terms of traveling and decreasing disasters. Several research on forecasting humidity over timeframes of minutes, days, months, and years have been undertaken in the last decade. Physical procedures, statistical or hybrid methods, such as neural networks, are the most often utilized strategies for estimating humidity day-ahead, according to a comprehensive set of forecasting methodology. The purpose of this research is to minimize prediction error. Using a recurrent neural network model and Deep Learning. Inthis paper proposed a three methodto evaluate the performance of relative humidity on the bases of deep learning algorithms. The process of prediction has been done based on these three models: Convolution Neural Network(CNN), Long-Short Term Memory (LSTM) model, and hybrid CNN-LSTM model. The researcher has used the real data of humidity for 30 years to train and testing the models.