Adaptação de um algoritmo de agrupamento para aplicação em dados de acidentes de trabalho

Abstract

The large number of accidents at work in Brazil leads the country to have a high cost with benefits, in addition to directly impacting the SUS (Unified Health System) and the quality of life of workers. The Public Ministry of Labor (MPT) launched the Occupational Safety and Health Observatory, where it is possible to access data on work accidents in Brazil in an organized manner.

The master's work Clustering Techniques applied to Work Accident Data (GIACOMELLI, 2020) aimed to apply data clustering algorithms in the work accident database to try to obtain useful patterns. This work supported the master's project through the development of techniques that allowed performing experiments with clustering algorithms in the accident database. For this, it was necessary to treat the database, adapt a hierarchical clustering algorithm to deal with a large database and implement a distance measure for categorical attributes.

This work adapted a software developed in the java language with implementations of new functionalities to treat the base, removing irrelevant attributes, objects with null information and performing transformations of some attributes. in the algorithm of grouping, new implementations were developed for the calculation of the distance measure between the objects and for which the large volume base was processed. Adaptations were also developed in the validation measure used. At the end of the process, the implementations allowed the execution of the algorithm based on work accidents with a satisfactory execution time.

The large number of accidents at work in Brazil leads the country to have a high cost with benefits, in addition to directly impacting the SUS (Unified Health System) and the quality of life of workers. The Public Ministry of Labor (MPT) launched the Occupational Safety and Health Observatory, where it is possible to access data on work accidents in Brazil in an organized manner.

The master's work Clustering Techniques applied to Work Accident Data (GIACOMELLI, 2020) aimed to apply data clustering algorithms in the work accident database to try to obtain useful patterns. This work supported the master's project through the development of techniques that allowed performing experiments with clustering algorithms in the accident database. For this, it was necessary to treat the database, adapt a hierarchical clustering algorithm to deal with a large database and implement a distance measure for categorical attributes.

This work adapted a software developed in the java language with implementations of new functionalities to treat the base, removing irrelevant attributes, objects with null information and performing transformations of some attributes. in the algorithm of grouping, new implementations were developed for the calculation of the distance measure between the objects and for which the large volume base was processed. Adaptations were also developed in the validation measure used. At the end of the process, the implementations allowed the execution of the algorithm based on work accidents with a satisfactory execution time.