Forecasting Analysis of Hourly Temperature using LSRU Based Hybrid Deep Learning Model

Over the years due to various human-made and natural disasters, there is a sharp change in the temperature pattern across the planet. A minimal sudden increase or decrease in the natural temperature pattern imposes a significant threat to the natural habitat, crop production, energy demand, etc. With the gradual shift of air temperature patterns in recent years, the precise forecasting of future temperature data is grabbing the attention of researchers. Due to the nonlinear essence of the temperature data, prediction models based on recurrent neural networks (RNNs) are commonly employed. In this work, a hybrid deep learning-based model named LSRU has been proposed which is a combination of Long Short Term Memory (LSTM) and Gated Recurrent Unit (GRU) models; to accurately forecast the hourly temperature. Here a detailed systematic study of all the commonly used RNN-based prediction models in the field of temperature prediction along with our proposed model is conducted. A complete ablation study has been conducted to explore the importance of the inclusion of the 1D-Convolutional Neural Network (CNN) layer in the prediction models. The experimental findings demonstrate the superior performance and stability of our proposed model compared to other state-of-the-art RNN models, across both short-term and long-term forecasting intervals.