Machine Learning Approach to Estimate Requirements for Target Productivity of Garments Employees

Data-driven approaches are widely used to improve the productivity of organizations. The availability of data related to business requirements and employee benefits even allowed us to predict productivity at the employee level. In our study, we investigated several machine learning approaches to predict productivity and estimation approaches to estimate team operating factors such as the amount of incentive and overtime that ensure a target productivity by the employees of a garment industry. This study used 3 months of data from a garment industry that recorded department ID, standard minute value, incentive, overtime, number of workers, and actual productivity. Our main contribution was to find the best estimation approach integrated with the best machine learning technique that recommends the required amount of incentive and overtime for a business to obtain a target productivity. For estimating business requirements, we explored the performance of Simulated Annealing, Particle Filter (PF), Fixed Point Particle Filter, and Particle Swarm Optimization algorithms. The PF displayed notable accuracy for incentive and overtime estimation, achieving the highest R-squared value of 0.60 for incentive estimation and 0.08 for overtime estimation. Additionally, the proposed machine learning model outperformed the existing Random Forest and Deep Learning model with a minimalist MAPE (Mean Absolute Percentage Error) of 6.5906. Our study concluded that the PF estimation algorithm has a strong potential to find the garment industry's business requirements.