The integration of machine learning for species identification is becoming increasingly important in entomological research. However, automatic species identification faces significant challenges such as low resolution, sample discoration, and small dataset sizes, which impede the reliability of traditional machine learning methods. Building upon the previous research on quantification of the color patterns of Stenaptinus occipitalis jessoensis using R-based analysis, this study demonstrates how to overcome these challenges in training machine learning for species identification. This approach allowed us to successfully classify geographic variations of S. occipitalis. Our results demonstrate the model's ability to identify these variations, despite the small size of the image datasets. This advancement shed some light on the potential of machine learning to identify morphological variation in highly polymorphic species.

• Dogyun Han(Applied Biology Program, Division of Bio-resource Sciences, Kangwon National University)
• Sam-Kyu Kim(Applied Biology Program, Division of Bio-resource Sciences, Kangwon National University, Department of Plant Medicine, Division of Bio-Resource Sciences, Kangwon National University)