Pietro Di Maggio
CIAD: Clustering-based IQR Anomaly Detector for Anomalous Weight Detection in Automotive Bill of Materials.
Rel. Luca Cagliero. Politecnico di Torino, Corso di laurea magistrale in Ingegneria Informatica (Computer Engineering), 2026
Abstract
Automotive Bills of Materials (BoMs) managed through Product Data Management platforms such as LIBRA record, among other attributes, the weight of every component in a vehicle. Errors in these weight values are common in industrial practice and difficult to detect, since a plausible weight for one type of component may be entirely anomalous for another: comparing weights across an entire BoM without regard to a component's functional category produces unreliable results. This thesis addresses the problem by first grouping components into functionally homogeneous families based on their textual descriptions, and then detecting anomalous weights within each family. An extensive experimental study compares several clustering strategies for short, domain-specific component descriptions, including centroid-based and density-based algorithms, a prototype-based approach exploiting the dominant token of each description, and hierarchical text representations.
This study leads to the adoption of DBSCAN combined with a keyword-based post-clustering refinement mechanism, which corrects and completes the density-based clustering by identifying the token that best represents each cluster
Relatori
Anno Accademico
Tipo di pubblicazione
Numero di pagine
Informazioni aggiuntive
Corso di laurea
Classe di laurea
Aziende collaboratrici
URI
![]() |
Modifica (riservato agli operatori) |
