Expressive and semantic description of human capital diversity: a machine learning approach

Human capital (HC) and its diversity are valuable assets of organizations, contributing to the longevity of companies and the success of startups. Recently, the industry has employed machine learning (ML) techniques in human resource management (HRM) to automate people selection, but considering mainly their individual technical skills. Conversely, to obtain better team performance and avoid social bias, complementary capacities and diversity of HC are desired. Analyzing the HC diversity imposes many challenges, including its semantic description. Studies in HRM propose to solve the identification of HC diversity using clustering methods, but analyzing their result may be impractical because of the large number of features of current datasets. This work presents an automatic pipeline that chooses the best clustering method, and semantically describes clusters of employees learned with the minimum of external input, leaving the decision-makers only with the task of providing the candidates' dataset and the number of desired partitions. We introduce a semantic descriptor operator $\star$ to describe the characteristics of clusters, filtering only those statistically relevant, allowing easy interpretation and potential information to assist decision-making. Finally, we validate our pipeline with a real dataset to present its semantic potential.