Root Cause Analysis of Incidents Using Text Clustering and Classification Algorithms

Abstract

The purpose of this study is to cluster the injury narratives to extract the root causes behind the accidents. Analysis is done on incident data collected from the database of an integrated steel plant. Key terms generated from the clustering of incident scenario help us in finding root causes of that particular incident. This study also proposed specific measures to the management that would improve the safety performance. This study uses text document clustering to discover the hidden factors and causes behind the incidents. Understanding previous accidents is necessary to avoid future accidents. However, for companies, management of large accident databases, and accurately classifying accident narratives are very challenging issues. Therefore, the aim of this study is to accurately classify accident reports using text classification approaches and evaluate their usefulness. The study used two machine learning (ML) algorithms, namely random forest (RF), and support vector machine (SVM) and found that SVM performed best in classifying the accident narratives. Further, SVM was experimented with different tokenization of the preprocessed narratives to get more reliable results.