Abstract
The challenge of managing household food waste has garnered significant attention due to its environmental and economic impacts. This study evaluates the performance of OpenAI’s GPT-4, Google’s Gemini, Mistral, and Anthropic Claude APIs in classifying food ingredients into food loss groups, providing compositional details, and suggesting relevant recipes. Using a validated dataset of common household ingredients, we evaluated the models for their accuracy in classification and the quality of generated content using automated metrics such as ROUGE and BLEU scores. The results indicate that Anthropic Claude achieved the highest classification accuracy at 75\%, followed by Mistral at 62.5\%, OpenAI’s GPT-4 at 56.25\%, and Google Gemini at 31.25\%. Furthermore, GPT-4 achieved the highest ROUGE and BLEU scores for composition and recipe suggestions, indicating its superior text generation capabilities. Statistical analysis, including one-way ANOVA and t-tests, revealed significant differences in the performance of the models. This study provides insights into the strengths and areas for improvement of these AI models in the context of sustainable food practices.
Presenters
Ezequiel SantosPhD Student, Faculty of Technology | Digital Games Development, IADE, Portugal
Details
Presentation Type
Paper Presentation in a Themed Session
Theme
Food Production and Sustainability
KEYWORDS
Food Waste Management, Household Food Waste, Artificial Intelligence