← Back to Publications List

Deep Learning Applications for Early Prediction of Natural Disasters in Bangladesh

Students & Supervisors

Student Authors
Md. Asif Mahmud Sezan
Bachelor of Science in Computer Science & Engineering, FST
Purnima Barua
Bachelor of Science in Computer Science & Engineering, FST
Anika Absar Orpi
Bachelor of Science in Computer Science & Engineering, FST
Md Jarif Hasan
Bachelor of Science in Computer Science & Engineering, FST
Afroza Akter Neha
Bachelor of Science in Computer Science & Engineering, FST
Supervisors
Md. Mortuza Ahmmed
Associate Professor, Faculty, FST

Abstract

Bangladesh is known for natural disasters like floods, cyclones and drought are frequent and they interfere with the agricultural production and put the national food security under threat. The timely and precise prediction of disasters is crucial in safeguarding crop and livestock as well as rural livelihood. This paper reveals the ways in which deep learning can be used to create early alerts and favor climate-adaptive farming in Bangladesh. The study relies on the 30-year dataset (1995-2024) of annual tallies of disasters, forecasting effectiveness, amount of training data, and responding time of the system. The interrelations among the data size, prediction accuracy as well as response speed were investigated using descriptive statistics and trend analysis. The findings show a strong improvement of model prediction accuracy over the time-span, based on secondary data of the World Bank Global Index, and increasing more than 25% of 1995 to almost 98% of 2024 with a concomitant growth in the amount of training data of less than 10 GB to around 100 GB. In addition, the response time of the system was reduced, which meant that it could give quicker and more reliable predictions. The occurrence of disasters has also increased significantly, and this has underscored the need to have advanced prediction systems to protect the farming property. Deep learning is especially important in helping the government of Bangladesh to reduce the effects of disasters and manage the food system; through predicting disasters earlier and more accurately, artificial intelligence-based models can minimize agricultural losses, build adaptive capacity, and improve food security and climate resilience in the long term.

Keywords

Deep Learning Disaster Prediction Climate Resilience Agriculture Food Security.

Publication Details

  • Type of Publication:
  • Conference Name: Gazipur Agricultural University International Conference (GAUIC 2025)
  • Date of Conference: 12/12/2025 - 12/12/2025
  • Venue: Gazipur Agricultural University, Gazipur, Bangladesh
  • Organizer: Gazipur Agricultural University
View on Publisher