Predicting Permanent Disability of Construction Workers Using Support Vector Machine

Koç K., Künkcü H., Gürgün A. P.

7th international Project and Construction Management Conference (IPCMC2022), İstanbul, Türkiye, 20 - 22 Ekim 2022, ss.1351-1359

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: İstanbul
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.1351-1359
  • Yıldız Teknik Üniversitesi Adresli: Evet

Özet

Several environmental uncertainties and dynamic nature of the construction industry render it one of the riskiest industries around the globe. Hence, occupational accidents are omnipresent in the industry, which requires a robust safety risk management plan to reduce the number and severity of construction accidents. This study develops a machine learning model to predict permanent disability status of construction workers. To achieve this objective, a dataset of construction accidents that occurred in Turkey was collected in the first step. Then several data preprocessing methods were adopted to prepare the data for prediction task. For the machine learning application, Support Vector Machine (SVM) was used. The findings show that the SVM yielded the recall values of 0.6988 for classifying permanent disabled workers. The results of the feature importance analysis also show that working days lost, type of injury, and daily wage were the most significant attributes in the model. The findings are expected to help construction professionals develop safety management strategies to minimize accidents that may result in permanent disability.