Orman Yangını Kurtarma Operasyonları İçin Twitter ve GPT Tabanlı Analiz Kullanan Yapay Zekâ Destekli Karar Destek Sistemi

Öz

Bu çalışma, Türkiye'deki orman yangınları sırasında acil kurtarma koordinasyonunu desteklemek amacıyla GPT-4 tabanlı anlamsal analiz ile gerçek zamanlı Twitter verilerini bütünleştiren yapay zekâ destekli Orman Yangını Kurtarma Karar Destek Sistemi'ni (FFR-DSS) sunmaktadır. Sistem, Muğla ve Antalya illerini etkileyen 2021 Türkiye orman yangınlarına ilişkin tweet verileri üzerinde uygulanmıştır. FFR-DSS, kriz odaklı sosyal medya içeriklerinden aciliyet düzeyi, duygusal ton, konum bilgisi ve eyleme geçirilebilirlik göstergelerini çıkarmak amacıyla istem mühendisliğiyle tasarlanmış büyük dil modeli (BDM) tabanlı bir analiz hattı kullanmaktadır. Kelime Torbası (BoW) ve TF-IDF gibi geleneksel doğal dil işleme yöntemleri, bağlamsal açıdan zengin ve duygusal olarak değişken sosyal medya mesajlarını yorumlamada sınırlı kalabilmektedir. Buna karşılık FFR-DSS, etiketli eğitim verisine ihtiyaç duymadan GPT-4'ün sıfır atışlı anlamsal akıl yürütme yeteneklerinden yararlanmaktadır. Sistemin coğrafi risk haritalama bileşeni, tweet düzeyindeki sonuçları ilçe bazında risk puanlarına dönüştürmekte ve dinamik ısı haritaları oluşturarak kaynak tahsisi ve önceliklendirme kararlarını desteklemektedir. Elle etiketlenmiş bir örneklem üzerinde yapılan değerlendirmede FFR-DSS, %92,4 sınıflandırma kesinliği ve %89,1 duyarlılık elde etmiş; %72,8 doğruluk sağlayan BoW+SVM temel yöntemini önemli ölçüde geride bırakmıştır. Bulgular, BDM destekli sosyal medya analitiğinin afet yönetiminde gerçek zamanlı durumsal farkındalık, kurtarma koordinasyonu ve karar desteği için etkili bir yaklaşım olduğunu göstermektedir.

Anahtar Kelimeler

• Acil durum müdahalesi
• Coğrafi zekâ
• GPT-4
• Karar destek sistemi
• Orman yangını kurtarma
• Twitter analitiği

AI-Powered Decision Support System for Forest Fire Rescue Using Twitter and GPT-Based Analysis

Öz

This study presents the Forest Fire Rescue Decision Support System (FFR-DSS), an AI-powered framework integrating GPT-4-based semantic analysis with real-time Twitter data to support emergency rescue coordination during forest fires in Turkey. Using tweets from the catastrophic 2021 wildfires affecting Muğla and Antalya, FFR-DSS employs prompt-engineered large language model (LLM) pipelines to extract urgency scores, emotional tone, geographical information, and actionability indicators from crisis-related social media content. Unlike traditional natural language processing (NLP) approaches such as Bag-of-Words (BoW) and TF-IDF, which may struggle with contextually rich and emotionally variable messages, the proposed framework leverages GPT-4's zero-shot semantic reasoning capabilities without requiring task-specific labeled training datasets. A geospatial risk mapping component further aggregates tweet-level outputs into district-level risk scores and dynamic heatmaps, enabling emergency teams to identify critical areas and prioritize resource allocation. Evaluation against a manually annotated sample demonstrates that FFR-DSS achieves 92.4% classification precision and 89.1% recall, substantially outperforming the BoW+SVM baseline with 72.8% accuracy. These results demonstrate the potential of LLM-augmented social media analytics to transform unstructured crisis communications into actionable intelligence, improving situational awareness, rescue coordination, and data-driven decision-making during large-scale natural disasters.In addition to its natural language understanding capabilities, FFR-DSS incorporates a geospatial mapping component to visualize and monitor critical rescue zones dynamically. By aligning semantic outputs with location data, the system helps authorities prioritize interventions in areas of greatest need. Empirical evaluations show that FFR-DSS achieves higher classification accuracy and interpretability than prior baseline methods, offering a more robust, transparent, and context-aware framework for emergency response. The results suggest that GPT-based systems represent a transformative step in disaster management technology, particularly in scenarios requiring rapid analysis and decision-making based on unstructured social media content.

Anahtar Kelimeler

• Decision support system
• Emergency response
• Forest fire rescue
• Geospatial intelligence
• GPT-4
• Twitter analytics

Kaynakça

1. AFAD. (2021). 2021 Yılı Orman Yangınları Değerlendirme Raporu. T.C. İçişleri Bakanlığı Afet ve Acil Durum Yönetimi Başkanlığı.
2. Alam, F., Ofli, F., and Imran, M. (2018). CrisisMMD: Multimodal Twitter datasets from natural disasters. In Proceedings of the 12th International AAAI Conference on Web and Social Media (ICWSM).
3. Ashktorab, Z., Brown, C., Nandi, M., and Culotta, A. (2014). Tweedr: Mining Twitter to inform disaster response. In Proceedings of the 11th International ISCRAM Conference.
4. Bilgili, E., Coskuner, K. A., Usta, Y., and Goltas, M. (2019). Forest fire occurrence modeling in Southwest Turkey using MaxEnt. iForest – Biogeosciences and Forestry, 17, 10–18. https://doi.org/10.3832/ifor4321-016
5. Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., … and Amodei, D. (2020). Language models are few-shot learners. Advances in Neural Information Processing Systems, 33, 1877–1901.
6. Castillo, C., Mendoza, M., and Poblete, B. (2011). Information credibility on Twitter. In Proceedings of the 20th International World Wide Web Conference (WWW 2011) (pp. 675–684). ACM.
7. Daşdemir, İ., Aydın, F., and Ertuğrul, M. (2021). Factors affecting the behavior of large forest fires in Turkey. Environmental Management, 67(1), 162–175. https://doi.org/10.1007/s00267-020-01389-z
8. Doguc, O. (2022). Twitter Verisi ile Doğal Afet Müdahale Süreci İçin Karar Destek Uygulaması. Afet ve Risk Dergisi, 5(2), 408–419. https://doi.org/10.35341/afet.1144350

9. EFFIS. (2021). European Forest Fire Information System – Turkey country report 2021. European Commission Joint Research Centre.
10. Ghazouani, D., Lancieri, L., Ounelli, H., and Jebari, C. (2019). Assessing socioeconomic status of Twitter users: A survey. In Proceedings of the International Conference on Recent Advances in Natural Language Processing (RANLP 2019) (pp. 388–398). INCOMA Ltd. https://doi.org/10.26615/978-954-452-056-4_046
11. Huang, Q., and Xiao, Y. (2015). Geographic situational awareness: Mining tweets for disaster preparedness, emergency response, impact, and recovery. ISPRS International Journal of Geo-Information, 4(3), 1549–1568. https://doi.org/10.3390/ijgi4031549
12. Imran, M., Elbassuoni, S., Castillo, C., Diaz, F., and Meier, P. (2013). Extracting information nuggets from disaster-related messages in social media. In Proceedings of the 10th International ISCRAM Conference.
13. Imran, M., Castillo, C., Diaz, F., and Vieweg, S. (2015). Processing social media messages in mass emergency: A survey. ACM Computing Surveys, 47(4), 1–38. https://doi.org/10.1145/2771588
14. Imran, M., Ofli, F., Caragea, D., and Torralba, A. (2020). Using AI and social media multimodal content for disaster response and management: Opportunities, challenges, and future directions. Information Processing and Management, 57(5), 102261.
15. Nenkova, A., and McKeown, K. (2011). Automatic summarization. Foundations and Trends in Information Retrieval, 5(2–3), 103–233.
16. OGM. (2021). Orman Yangınları İstatistikleri 2021. Orman Genel Müdürlüğü, T.C. Tarım ve Orman Bakanlığı.
17. OpenAI. (2023). GPT-4 technical report. arXiv preprint arXiv:2303.08774.
18. Verma, S., Vieweg, S., Corvey, W., Palen, L., Martin, J. H., Palmer, M., Schram, A., and Anderson, K. M. (2011). Natural language processing to the rescue? Extracting “situational awareness” tweets during mass emergency. Proceedings of the International AAAI Conference on Web and Social Media, 5(1), 385–392. https://doi.org/10.1609/icwsm.v5i1.14119
19. Vieweg, S., Hughes, A. L., Starbird, K., and Palen, L. (2010). Microblogging during two natural hazards events: What Twitter may contribute to situational awareness. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI 2010) (pp. 1079–1088). ACM.
20. Ziems, C., Held, W., Shaikh, O., Chen, J., Zhang, Z., and Yang, D. (2023). Can large language models transform computational social science? Computational Linguistics, 50, 237–291. https://doi.org/10.48550/arXiv.2305.03514
21. Zou, L., Lam, N. S., Shams, S., Cai, H., Meyer, M. A., Yang, S., … and Reams, M. A. (2018). Social and geographical disparities in Twitter use during Hurricane Harvey. International Journal of Digital Earth, 12(11), 1300–1318.