Developing a Model for Product Quality Prediction and Improvement Using the Decision Tree Algorithm and Data Envelopment Analysis – a Case Study: Manufacturers of the Tiba Anti-roll Bar in Iran

Main Article Content

Nadereh Sadat Rastghalam, Roya Mohammad Alipour Ahari, Ahmad Reza Shekarchizadeh, Atefeh Amindost


Parallel to technological advancements, industries are paying closer attention to cost-effectiveness and product quality as prerequisites of good performance in a competitive world. Using the tools of artificial intelligence and machine learning can cut down on costs and wastes while increasing product quality. The present study employed the decision tree algorithm to explore waste patterns in the production line of anti-roll bars in Tiba. To do this, first, a database comprised 4169 pieces in the form of nine characteristics, and then one class was formed. After determining the patterns, the rules were evaluated via data envelopment analysis. By exploring the most important rules and finding solutions to eliminate them, the quality of the final product was foreseeably enhanced, consequently reducing the percentage of waste pieces and reworking. The proposed approach is recommended for companies with a high rate of waste and different working stations. According to the results, the most important criteria affecting the breakdown include the quality of cooling and soldering. In addition, the accuracy of the C5 algorithm was 94% in predicting the piece quality. The present study evaluated four rules at a depth of 1, 12 rules at a depth of 2, and 8 rules at a depth of 3. By exploring and evaluating such rules, the waste and reworking reduced while the quality increased. The model was validated by implementing reforming priorities in rolling and soldering from April to September 2021, suggesting that when the conditions were sustainable and the quality of input materials as well as other variables remained unchanged in 2021, product quality improved by 7%. This shows that the model is appropriately valid. 


Article Details