Thanks to funding from the Tuscany Region under the New Development and Cohesion Plan – FSC, supporting collaborative activities by public-private partnerships to develop and promote high-tech sectors in the region, the Department of Mathematics announces the launch of the MAmMUT project. This initiative aims to develop advanced machine learning methods for multi-type datasets. Below is a brief description of the project.
A critical challenge in the manufacturing sector is the ability to predict failures, the need for interventions, or critical conditions in general. Anticipating such events is essential for taking proactive actions and decisions to avoid or address these critical situations. These predictions rely on diverse types of data—numerical, categorical, or more complex formats like images—collected from various sources (sensors, RFID, NFC, external data, production monitoring systems, etc.) during different stages of the production process, particularly in manufacturing. These data are processed using mathematical and statistical algorithms within the framework of the Industry 4.0 paradigm. However, in the current landscape of predictive models, different methods may perform optimally for specific types of data but only sub-optimally for others. The core issue of the project is integrating different types of data into a single predictive model to enhance predictive capabilities while streamlining data collection and management. Throughout the project, various strategies will be explored to address this problem, potentially resulting in new types of models. The aim is to identify the most effective methods by testing them on case studies relevant to the manufacturing sector. The project’s impact extends beyond economic benefits, such as improved predictive capabilities, to fostering economic and environmental sustainability. This is achieved through opportunities for reusing and integrating collected data.