The main function of packaging is to protect and/or extend the life of the items they contain (by making storage and transit easier). Packaging assists with making storage, transit, and marketing of the product significantly easier. People found that transferring products from the point of production to the place of use was necessary as they expanded beyond small-group self-sufficiency. Products exposed to improper handling, storage, and transportation conditions resulted in damaged items. The temporary intermediate step of packaging helps to guarantee that products are received in good condition. It might be challenging to draw a line between the package and the product itself for some industries, such as sterile medical equipment. Imagine a liquid product like a detergent. How does the producer geographically position the product inside a market after the production process has ended without the package? Although you might be willing to restrict sales of your product to your local neighborhood, selling all your products inside the manufacturing facility may be problematic. One of the most innovative technological advancements of the twenty-first century is advanced data analytics, which allows the uncovering of underlying trends through powerful computing techniques. Millions of online product reviews are published by customers on various e-commerce platforms, and these reviews may provide designers with priceless information about how to design products. In this research, we developed a web-based intelligent packaging evaluation approach to evaluate packaging design and problems based on Customer reviews on the E-Commerce platform. From the packaging aspect, the protection function of the product's packaging is essential. The evaluation process before and after design implementation lets the designer know how to improve packaging features. These evaluations are possible with a series of physical tests. If the package is after design implementation, redesigning and redoing all physical tests is expensive and time-consuming. Therefore, an alternative evaluation method is needed to extract packaging defects and find causes of damage without redoing physical tests. Traditional methods used surveys and focus group interviews to extract information from customers' voices. Since these methods were applied to a small group of customers and needed interpretation by a human, they were inaccurate, time-consuming, and expensive. In this research, we used machine learning algorithms: LSTM, Naive Bayes classifier for natural language processing and Fp-Growth for association rule mining to determine customers' attitudes (positive, negative, and neutral) and the relationship between critical damaged features of products or packaging. The objective is to utilize this application's data output to identify a series of packaging issues based on customer reviews. The web-based application aims to refine a large volume of qualitative data into quantitative data on product features and packaging issues via sentiment analysis and find the most critical relationship between issues/features mentioned in online reviews based on association rule mining. The relationship between negative words is important because the sentiment analysis score cannot provide a detailed context of what causes the sentiment and how they cause it. As a result, designers can make more informed decisions to improve the packaging manufacturing process or packaging selection for the product. For case studies, we chose two detergent designs: liquid bottles and pod detergents. Because this product has different container and packaging designs, there are some negative complaints about their design and a large volume of reviews with various positive and negative reviews available. As a result, we compared the packaging failure of these two packaging designs and identified the most critical problem with each design.
Read
