Application of optimized machine learning techniques for prediction of occupational accidents

•Both categorical data and unstructured text data have been analysed.•Incident outcomes are predicted using GA and PSO optimized SVM and ANN approaches.•PSO-SVM outperforms other algorithms in terms of accuracy (i.e., 90.67%).•Nine useful rules are extracted using PSO-SVM based C5.0 algorithm.•Root causes of incidents have been identified.