UTILIZATION MACHINE LEARNING TO BUILD A PREDICTIVE MODEL TO IMPROVE PACKAGING QUALITY IN PRODUCTION PROCESSES

Abstract

The study is geared toward the development of a predictive model of packaging quality since it plays an important role in the food industry. Nutritional value losses and spoilage are directly associated with packaging defects. The study factors those variables that strongly influence the quality of packaging and attempts early prediction of defectiveness, hence a reduction in financial loss, which will drive continuous improvement efforts to make better quality management decisions. The fruits of data mining and the machine learning revolution have essentially made effective predictive methods possible. An applied analytical approach based on previous studies and production records from Noor Al-Kafeel Company's chicken and chicken parts factories was adopted by this research for constructing the proposed model. Methodology used machine learning techniques to analyze relationships between a set of operational, human, and environmental variables that may predict packaging quality. 300 real-world operational observations suitable for modeling comprised the sample. Data mining techniques and the Random Forest Algorithm were used. The predictive model produced an accuracy rate of 97.6%, hence proving robustness and applicability in industrial environments . Factors that influence packaging quality analyzed included defect rate, humidity, heat sealing, and machine speed; a revelation that these factors ranked high in terms of their importance. Due to automation, the human factor is less significant. These findings make it possible for the proposed model to help in moving from post-inspection detection of defects to proactive, predictive methodology, which would improve production process efficiency, add value and sustain industrial organizations.