This study aims to propose new grading standards that can be applied to AI-based automatic sorting machines, reflecting current distribution and consumption trends. The current domestic grading standards for onions in South Korea are based on the “Agricultural and Fishery Products Quality Control Act”. They classify onions based on criteria such as uniformity, shape, color, and the presence of foreign matter. Onion grading standards are divided into four categories based on bulb diameter and weight. However, in the actual domestic market, onions are distributed according to a five-grade classification based on bulb diameter. Therefore, this study classified onions into eight grades, reflecting current distribution and consumption trends in the domestic market. These grades are applicable to AI-based automatic sorting machines. Marketable onions were classified into A1 (extra large) to A5 (extra small) based on the diameter of a single bulb. Onions used for non-marketable purposes (processing) were classified as grade B. Additionally, grade C and grade D were designated for processing and disposal, respectively. By establishing quality grading classifications that align with current distribution and consumption market trends as well as the operational characteristics of AI-based automatic sorting machines, we can expect improvements in work efficiency and reductions in distribution costs. Following this study, it will be necessary to establish comprehensive quality grading standards that include both external criteria (such as bulb weight and size) and internal criteria (such as detection of internal decay and disease occurrence).

This study proposes a weight optimization technique based on Mixture Design of Experiments (MD) to overcome the limitations of traditional ensemble learning and achieve optimal predictive performance with minimal experimentation. Traditional ensemble learning combines the predictions of multiple base models through a meta-model to generate a final prediction but has limitations in systematically optimizing the combination of base model performances. In this research, MD is applied to efficiently adjust the weights of each base model, constructing an optimized ensemble model tailored to the characteristics of the data. An evaluation of this technique across various industrial datasets confirms that the optimized ensemble model proposed in this study achieves higher predictive performance than traditional models in terms of F1-Score and accuracy. This method provides a foundation for enhancing real-time analysis and prediction reliability in data-driven decision-making systems across diverse fields such as manufacturing, fraud detection, and medical diagnostics.