Deep learning for big weather data analyzing and forecasting

Document Type : Research Paper


1 Department of Economics, College of Economics and Administration, Duhok University, Duhok, Kurdistan Region, Iraq

2 Department of Computer Science, Faculty of Science, Zakho University, Duhok, Kurdistan Region, Iraq



Weather prediction is vital in daily life routines, for risk mitigation and resource management such as flood
risk forecasting. Quantitative prediction of weather changes depends on different parameters such as rainfall time,
temporal, barometric pressure, humidity, precipitation, solar radiation and wind. Therefore, a highly accurate system
or a model to forecast the highly nonlinear changing happening in the climate is required. The focus of this research
is direct prediction of forecasting from weather-changing parameters, the forecasts are performed using collected data
values recorded in a big dataset (the dataset collects the weather parameter data of the Canary Islands (Las Palmas,
Tenerife a Palma, Fuerteventura, La Gomera, Lanzarote and Hierro). The forecasting system is performed by proposing
a deep learning approach (CNN). The research goal is predication the weather condition. The acquired classification
accuracy for the climate condition using CNN (ShuffleNet) structure is 98%, and the recall and Precision results are 97.5
and 96.9 respectively


Articles in Press, Corrected Proof
Available Online from 23 March 2023
  • Receive Date: 04 January 2023
  • Revise Date: 20 February 2023
  • Accept Date: 11 March 2023
  • First Publish Date: 23 March 2023