Hybrid Predictive Model for Efficiency Forecasting in Garment Production: Integrating Ensemble and Time-Series Approaches

Sri Lanka's garment industry is a critical component of the nation's economy, contributing significantly to GDP, export revenue, and employment, particularly for rural women. While the industry is recognized globally for its ethical production and high-quality outputs, it faces challenges including dependence on imported raw materials, fluctuating global demands, and competition from low-cost manufacturing hubs like Bangladesh and Vietnam. To maintain its competitive edge, the industry must explore innovative solutions such as predictive analytics and advanced modeling techniques to optimize efficiency and sustainability. This research examines the application of predictive analytics and time series modeling to enhance production efficiency in Sri Lanka's garment sector. We explore various machine learning models, including ensemble techniques and Long Short-Term Memory (LSTM) networks, to forecast production efficiency. Using data from a major Sri Lankan garment manufacturer, spanning May 2023 to August 2024, we develop a hybrid predictive model. Data preprocessing involved steps like cleaning, handling missing values, and feature engineering, ensuring reliable inputs for the model. Our results show that the hybrid model provides a balanced approach, offering practical advantages in real-world production environments. The model demonstrates moderate prediction accuracy, reducing mean squared error (MSE) and mean absolute error (MAE), outperforming several other traditional models.