Field lab Jerónimo Martins: optimization of retail operations - a semi-supervised learning approach for anomaly detection in lot ovens

In this work project,we demonstrate how machine learning can be used to analyze Internet of Things (IoT) data to monitor food preparation processes and detect faults. The developed models have comparable performances to humans. Additionally, we show that semi-supervised learning can significantly increase the fault detection model’s performance. It does so by projecting labels to prior unlabelled observations through label spreading. Results from the developed model suggest a large potential for machine learning based processes with less need for human intervention