Climate Forecasting in Sulaymaniyah City Using Deep Learning Techniques
Volume 11 , Issue 3 , August 2025 , Pages 54-74
Authors
Rebin Abdulkareem Hamaamin 1 ; Omar Mohammed Amin Ali 2 ; Shahab Wahhab Kareem 3
1 Charmo University- Computer department- Sulaymaniyah - Iraq
2 Department of Information Technology, Chamchamal Technical Institute, Sulaimani Polytechnic University, Sulaymaniyah, Iraq
3 Department of Technical Information Systems Engineering, Technical Engineering College, Erbil Polytechnic University, Erbil, Iraq
10.17656/sjes.10196
Keywords
Abstract
Climate forecasting has grown tremendously in recent years and is expected to grow further in the coming years. In the previous decade, extensive research has been carried out on weather forecasting at different time intervals in minutes, hours, days, months, and years. This research uses a neural network model to categorize and predict the weather in Sulaymaniyah, which is located in the Kurdistan Region of Iraq. The use of neural network models for climate data analysis has advanced. With the help of another parameter that affected the prediction, the dataset-based forecasting approach uses a single-column time series to depict the future. A real-time weather station dataset from the Sulaimani Meteorological and Seismological Directorate is used in the implementation model. The data collection includes implementation data from the weather station for an iterative neural network model that forecasts future climate and displays historical propagation. From 1993 to 2023, the dataset includes daily data on average temperature, humidity percentage, and precipitation for each of the twelve months from January to December. It was gathered every day for thirty years and includes information about Suleimani City's maximum and minimum temperatures, average temperatures, humidity levels, and rainfall over that time. One of the other goals of this investigation is to test and compare the models used to predict relative average temperatures. As the results indicate that Bi-LSTM and GRU outperformed GBT in both training and testing. The study also suggests areas for further investigation into integrated approaches to assist researchers in the field in creating more accurate forecasting systems.