Industry 4.0 is a process that is leading to fully automated and interconnected industrial production. A guideline of this phenomenon is constituted by analytics: once the data has been collected, it is necessary to derive value from it. Even in the industrial field, companies are trying to exploit the large amount of data that follows. In this context, machine learning topics provide a tool of indisputable relevance in a variety of applications. This thesis deals with the development of a feature-based anomaly detection method related to a case study of a company in the entertainment field. Specifically, motivated by the lack of a reliable labeled set, the approach takes shape in an unsupervised scenario. In synergy with this, a tool is adopted that provides the interpretability of the results. Understanding why a point is labeled anomalous is becoming increasingly important, especially by virtue of root cause analysis.
L’industria 4.0 è un processo che sta portando alla produzione industriale del tutto automatizzata e interconnessa. Una direttrice di questo fenomeno è costituita dagli analytics: una volta raccolti i dati, bisogna ricavarne valore. Anche in ambito industriale le aziende stanno cercando di valorizzare la grande mole di dati che ne consegue. In questo contesto, i temi dell’apprendimento automatico forniscono uno strumento di indiscutibile rilevanza in svariate applicazioni. Questa tesi tratta l’utilizzo di un metodo di riconoscimento delle anomalie relativo ad un caso studio di un’ azienda del settore dell’intrattenimento. Nello specifico si tratta di un contesto non supervisionato in quanto l’etichettatura dei dati non è disponibile a priori. In sinergia a ciò viene adoperato uno strumento che fornisca un’interpretabilità ai risultati ottenuti in modo da fornire un aiuto nell’analisi delle cause principali.
Rilevamento anomalie per macchine automatiche dell'industria dell'intrattenimento
PERONI, MARCO
2021/2022
Abstract
Industry 4.0 is a process that is leading to fully automated and interconnected industrial production. A guideline of this phenomenon is constituted by analytics: once the data has been collected, it is necessary to derive value from it. Even in the industrial field, companies are trying to exploit the large amount of data that follows. In this context, machine learning topics provide a tool of indisputable relevance in a variety of applications. This thesis deals with the development of a feature-based anomaly detection method related to a case study of a company in the entertainment field. Specifically, motivated by the lack of a reliable labeled set, the approach takes shape in an unsupervised scenario. In synergy with this, a tool is adopted that provides the interpretability of the results. Understanding why a point is labeled anomalous is becoming increasingly important, especially by virtue of root cause analysis.File | Dimensione | Formato | |
---|---|---|---|
Peroni_Marco.pdf
Open Access dal 15/10/2023
Dimensione
6.56 MB
Formato
Adobe PDF
|
6.56 MB | Adobe PDF | Visualizza/Apri |
The text of this website © Università degli studi di Padova. Full Text are published under a non-exclusive license. Metadata are under a CC0 License
https://hdl.handle.net/20.500.12608/29244