Ebrahim Rajabpour

Background: In today’s world, where organizations operate in complex and competitive environments, accurate employee performance evaluation has gained increasing importance as a key component of human resource management. In the gas industry—where sustainability, efficiency, and data-driven decision-making are of critical significance—reliance on traditional performance appraisal methods may involve challenges such as human error, bias, and lack of transparency. The use of modern technologies such as machine learning opens new horizons for improving performance evaluation processes. Objective: This study aims to design and implement a scientific, data-driven model for evaluating the performance of employees at the Bushehr Gas Company. Utilizing the decision tree algorithm as an interpretable and precise method, the research attempts to classify and analyze employee performance with high accuracy and transparency. The main focus is on identifying key indicators and examining their impact on predicting desirable or improvable performance. Methodology: First, through a systematic review of theoretical sources and the application of the Delphi method among eight experts from the Gas Company and Persian Gulf University, a final set of 20 indicators for employee performance evaluation was extracted. Performance data for 50 employees over 24 time periods were collected. Data analysis was performed using Python and the decision tree algorithm (with Entropy criterion and a depth of 5). The model was evaluated by splitting the dataset into training and test sets, and outputs were assessed in terms of accuracy, recall, precision, and F1-score. Findings: The results indicated that the designed algorithm successfully classified employees into two groups—“fit for position” and “in need of improvement”—with 100% accuracy. Indicators such as “number of tasks completed,” “creativity,” “learning ability,” and “colleague interaction” had the greatest influence in determining performan