Forecasting Daily Relative Temperature in Hmawbi, Myanmar Using Convolutional Neural Network and Deep Learning Models

Relative temperature (maximum and minimum) is a significant factor influencing climate change and global warming, and it has become an important research focus in recent decades. This study aims to apply a deep learning approach using a Convolutional Neural Network (CNN) model to forecast daily relative temperature in Hmawbi Township, Myanmar. The model is developed using average daily relative temperature data collected from 2013 to 2022. Based on the forecasting results, a slightly decreasing trend in relative temperature is observed. The CNN model is employed to learn the underlying patterns in historical temperature data collected from 2013 to 2017 and to generate forecasts for the period 2018–2022. In this paper, two prediction models are developed by using Convolutional Neural Network (CNN): one with the original data (without noise removal), one with data cleaned using the Simple Moving Average method (SMA). The performance and accuracy of the CNN model are evaluated using root mean squared error (RMSE) and mean square error (MSE). The use of CNN allows for effective feature extraction and improved handling of nonlinear relationships in the dataset. CNN achieves RMSE of 5.398 and MSE of 29.020, SMA achieves RMSE of 3.767 and MSE of 14.841. The results indicate that the CNN with SMA deep learning model achieves lower error values, demonstrating higher prediction accuracy and better performance compared to traditional statistical methods. The CNN-based deep learning model is found to be a reliable and effective approach for forecasting relative temperature in Hmawbi Township.