Analysis and Classification of Civil Servant (CPNS) Needs Using the Random Forest Method

Abstract

Synergy between the State Civil Service Agency (BKN) and the Ministry of Administrative and Bureaucratic Reform (PANRB) is crucial in coordinating the planning and procurement of State Civil Apparatus (ASN) to ensure optimal public services. However, the complexity of mapping new positions often leads to bureaucratic inefficiencies due to the lack of accuracy in determining job qualifications and setting compensation standards. This study aims to develop and test the reliability of an algorithm-based predictive model.Multi-Output Random Forest Classifieras a decision support system (data-driven decision-making) in national personnel regulation planning. Experiments and model testing were conducted using a large-scale real dataset of civil service procurement sourced from the sscasn_formasi_cpns_2024.xlsx document. The operational input variables tested included agency name, position type, placement location, and formation allocation quota. Simultaneously (multi-output), the model was trained to predict two target variables simultaneously, namely Education Level (High School, D3, S1) and Compensation Level (Low, Middle, High). Model performance testing was evaluated through amajority votingfrom a set of decision trees (decision trees) by minimizing the Gini impurity index (Gini Impurity). In order to eliminate the natureblack-boxIn artificial intelligence, the level of model transparency is strengthened by feature importance value analysis (Feature Importance) through the approach Mean Decrease Impurity(MDI). The results of the model test prove that the integrationmachine learningable to objectively identify non-linear relationship patterns of heterogeneous formation characteristics. This automated approach has strategic implications for planning agencies in establishing appropriate graduation qualifications and minimizing human bias in standardizing public sector job classes.