SIGMUND: Optimization of DISC Methodology and distribution of groups with Machine Learning

Proceedings of The 5th International Conference on Modern Research in Engineering, Technology and Science

Year: 2022

DOI: https://www.doi.org/10.33422/5th.icmets.2022.02.15

[Fulltext]

SIGMUND: Optimization of DISC Methodology and distribution of groups with Machine Learning

Cleiton Silva Ribeiro, Higor Barbosa Ferreira, João Pedro Rodrigues, Ricarte Elias, Victor Santos Boroni, Rafaela Priscila Cruz Moreira, Luiz Melk de Carvalho, Flávio Henrique Batista de Souza

ABSTRACT: 

In this work, the SIGMUND software is presented and an optimization process for the behavioral assessment questionnaire used in DISC (Dominance, Influence, Steadiness, Compliance) methodology is described. DISC is a tool that sets out to establish objectives aligned with the professional profile of employees by measuring strengths and weaknesses to achieve better results. This questionnaire is typically expensive because it covers several aspects through a broad variety of questions to assess the professional profile. Therefore, in this work, the Bagged Decision Trees (BDT) algorithm was implemented to reduce the number of questions without losing the quality of the test. The BDT estimates the importance of attributes within a database, returning a score for each attribute. In this algorithm, the higher the score, the greater the importance. After optimizing the questionnaire, it was used to define the profile of candidates and, subsequently, it was created academic groups that allow for better interaction and experience between members. As these groups are created randomly or by free choice, there may be conflicts or even not taking full advantage of the contribution that each one would have if the profile of these members were taken into account. For the creation of these smart groups based on the profile of the candidate, the K-means clustering algorithm was applied to define an ideal number of people in each group to guarantee that there is a balance. As a result, the BDT managed to reduce the number of questions by 52% with an accuracy level of 75.8% and for the division of groups in equal variation, the K-means obtained an accuracy of 97%.

keywords: Academic Groups; Behavioral Assessment; Data Mining; K-Means Algorithm; Quiz .