The event, that will be held in Vilnius (Lithuania) from September 9th to 13th 2024, is one of the most renowned European conferences dedicated to machine learning, attracting international experts to discuss the latest innovations and trends in the field.
Our contribution focuses on the application of Graph Neural Networks (GNN), advanced machine learning models based on graphs, to optimize new product development activities. The case presented focuses on the food industry but it is representative of the needs also felt by other sectors, such as cosmetics and chemicals.
The paper, titled “Graph Machine Learning for Fast Product Development from Formulation Trials,” was jointly developed by Intellico and the Department of Computer Science at the University of Milan and will be presented at the “European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases” – ECML PKDD – 2024.
The paper falls under the category of “Applied Data Science“, which aims to bridge the gap between theory and practice, showing how technological solutions can be effectively implemented in practical contexts to generate a tangible impact.
To test the model, Perfetti Van Melle, a leading multinational company in the confectionery sector and working with Intellico on the introduction of artificial intelligence in R&D and new product development processes, was involved.
Our focus: Explainable AI for Rapid New Product Development
The paper, titled “Graph Machine Learning for Fast Product Development from Formulation Trials,” proposes an innovative technique to speed up new product development using advanced graph-based machine learning models.
Creating and introducing new products to the market, with significant improvements, requires multiple formulation trials, and exploring numerous variables and properties. The traditional approach involves lengthy and iterative experiments based on “trial and error” approach, which can be inefficient and expensive. Machine learning offers a promising path to streamline this process, enhancing efficiency and innovation. However, the lack of interpretability and explainability of the models poses significant limitations in adopting these technologies.
In our work, we propose a methodology for rapid new product development using graph-based machine learning models and explainability techniques. Starting from tabular formulation data, the model learns a latent graph among the elements and predicts the properties of a formulation. Subsequently, Explainable AI supports the user, and the R&D team, in identifying new recipes to achieve desired properties.
We evaluated the model on two datasets: one related to the development of new food formulations/recipes and another in an alternative context (healthcare sector benchmark). The results demonstrate the model’s effectiveness in predicting the outcomes of new formulations, drastically reducing the need for physical experiments and material waste. This methodology not only speeds up the development process but also promotes more sustainable resource management.
Admission to the Conference is a recognition of the results achieved by Intellico, in collaboration with prominent academic institutions such as the Department of Computer Science at the University of Milan, in the field of Explainable AI. It is also a recognition to Intellico’s commitment in providing a “practical ground” for research to test new technological advancements/trends in the field.
Conclusions
Participation in the ECML PKDD 2024 represents an important opportunity for Intellico to share the innovations developed in collaboration with prominent academic institutions in the field of research, such as the Department of Computer Science at the University of Milan, and obtain valuable feedback from the international scientific community. We are proud to contribute to the growth and evolution of the field of machine learning and artificial intelligence with solutions that have a tangible and sustainable impact. Our presence at this conference further strengthens our commitment to staying at the forefront of technological innovation.
A special thanks to the team from the Department of Computer Science at the University of Milan and the R&D team of Perfetti Van Melle for their collaboration and the incredible energy dedicated to research.
If you are interested in learning more about Intellico’s applications in the sector of Consumer Packaged Goods https://intellico.ai/blog/explainable-ai-what-is-it-about-lets-discover-it-in-the-cpg-sector/
Contributors
Raffaele Olmeda, Head of Data Science Department Intellico.ai
Margherita Pindar, Intellico.ai Developer
Matteo Zignani – Associate Professor of Computer Science at the University of Milan
Manuel Dileo – PhD candidate in computer science at the University of Milan
