Deep Q-Network-Based Agent Learning for Employee Performance Prediction in Industrial Environments

Employee performance prediction is essential for workforce optimization in industrial environments, enabling organizations to enhance productivity, balance workload distribution, and improve decision-making. Traditional human resource management (HRM) approaches often lack the adaptability required for dynamic workplace conditions. This study proposes a Deep Q-Network (DQN)-based reinforcement learning model for predicting employee performance and optimizing task allocation. The model employs a multi-objective reward function to balance task complexity, workload fairness, employee experience, and efficiency while minimizing workforce imbalances. Experimental results demonstrate that the DQN-based agent effectively improves task assignment strategies, increasing cumulative rewards over training episodes. The model exhibited strong learning capabilities, as shown by an upward trajectory in Q-values and convergence of the loss function. The task allocation process ensured optimal utilization of employee capabilities, with the model consistently identifying high-performing employees and assigning tasks accordingly. The model’s allocation strategy effectively identified and prioritized high-performing employees, ensuring efficient task distribution. Further refinements in the reward function can enhance fairness by improving workload balance while maintaining productivity.