Modelo de predicción de riesgos psicosociales en el transporte urbano de pasajeros usando técnicas de Inteligencia Artificial

Abstract

EXECUTIVE SUMMARY The city bus driver job ranks among the highest risk and most stressful modern occupations. Modern technologies provides greater autonomy and work flexibility, however they also expose drivers to psychosocial risks, which leads to work stress. Consequently, the early prediction of stress and their associated risk, would contribute to make preventive decisions. The objective of this study is to develop a model that allows predicting psychosocial risks in urban passenger transport in the city of Ambato, applying supervised machine learning techniques. For this purpose, we used data set of occupational psychosocial risk of urban bus drivers obtained with the Fpsico 4.0 questionnaire. The study applies the methodology for the identification, analysis, and evaluation of psychosocial risks of the INSHT of Spain and the Cross Industry Standard Process for Data Mining (CRISP-DM) framework. The classification is performed with the three non-parametric supervised algorithms: k-nearest neighbors, decision trees and support vector machine. The evaluation metrics of the algorithms used are the Jaccard index and F1-score. The experimental results show that the support vector machine model shows better performance with an F1 score of 93 percent and the Jaccard score of 87 percent.