Developing a Model for Product Quality Prediction and Improvement Using the Decision Tree Algorithm and Data Envelopment Analysis

In today’s competitive industrial landscape, technological progress has made cost efficiency and product quality critical for optimal performance. Artificial intelligence (AI) and machine learning (ML) offer powerful tools to minimize waste, reduce costs, and enhance product standards. This study leverages the decision tree algorithm (C5) to identify waste patterns in the production line of anti-roll bars at Tiba, analyzing a dataset of 4,169 units with nine features and one class label. The derived patterns were then assessed using data envelopment analysis (DEA) to prioritize actionable rules for quality improvement. Key findings revealed that cooling and soldering quality were the most influential factors in production defects, with the C5 algorithm achieving 94% accuracy in predicting product quality. The study evaluated 4 rules at depth 1, 12 at depth 2, and 8 at depth 3, demonstrating how targeted interventions based on these rules reduced waste and rework. Practical validation was conducted by reforming rolling and soldering processes between April and September 2021. Under stable input conditions, these adjustments led to a 7% improvement in product quality, confirming the model’s validity. This approach is particularly recommended for manufacturers with high waste rates and multi-stage production lines, offering a data-driven pathway to operational efficiency and superior output.