Wipe : A Web-Based Intelligent Packaging Evaluation Platform Using Machine Learning and Association Rule Mining

This paper aims to propose a web-based intelligent packaging evaluation (WIPE) platform as an alternative method for assessing the performance of product and packaging systems during e-commerce distribution. The packaging industry primarily relies on laboratory evaluations to evaluate packaging performance under controlled conditions. Still, e-commerce distribution presents unique challenges due to increased handling points and unexpected hazards that are not always captured in physical tests. WIPE uses machine learning algorithms and association rule mining to extract packaging defects and identify causes of damage based on customer reviews on a real e-commerce platform. The platform correlates images and text of online product reviews and determines relationships between the most frequent words in customer reviews to predict damages and their causes and effects. To evaluate the WIPE platform's performance, two case studies focusing on laundry detergent liquid bottles and pods on the Amazon platform were conducted. Results demonstrated that WIPE was enable to extract relevant information from customer reviews, identify packaging defects, and predict potential causes of damage. The research contributed to the field of packaging evaluation by achieving automation of packaging evaluation based on customer reviews. Additionally, it introduced a novel method for correlating images and text of online product reviews and brought sentiment analysis and association rule mining to the packaging evaluation area. The WIPE platform potentially enhanced product and packaging design and ultimately improved customer satisfaction, particularly for e-commerce distribution