Machine Learning Based a Comparative Analysis for Detecting Tweets of Earthquake Victims Asking for Help in The 2023 Turkey-Syria Earthquake

Authors : Anıl Utku, Ümit Can

Pages : 55-62

Doi:10.55195/jscai.1365639

View : 72 | Download : 89

Publication Date : 2024-01-06

Article Type : Research Paper

Abstract :Two major earthquakes in Kahramanmaraş on February 6, 2023, 9 hours apart, affected many countries, especially Turkey and Syria. It caused the death and injury of thousands of people. Earthquake survivors shared their help on social media after the earthquake. While people under the rubble shared some posts, some were for living materials. There were also posts unrelated to the earthquake. It is essential to analyze social media shares to plan the process management effectively, save time, and reach the victims as soon as possible. For this reason, about 500 tweets about the 2023 Turkey-Syria earthquake were analyzed in this study. The tweets were classified according to their content as user tweets under debris and user tweets requesting life material. Popular machine learning methods such as DT, kNN, LR, MNB, RF, SVM, and XGBoost were compared in detail. Experimental results showed that RF has over 99% classification accuracy.
Keywords : Earthquake, Machine learning, Random forest, Sentiment analysis