Μελέτη των ναυτικών ατυχημάτων με χρήση στατιστικών τεχνικών και μηχανικής μάθησης
View/
Keywords
Ναυτικά ατυχήματα ; Μηχανική μάθηση ; Στατιστικές τεχνικές ; Λογιστική παλινδρόμηση ; Δέντρα απόφασης ; ΣυσταδοποίησηAbstract
This thesis focuses on predicting the occurence of fatalities in maritime accident incidents,
utilizing data from the Transportation Safety Board of Canada (TSB).The methodology is based
on logistic regression, decision trees (CART), and clustering. Variable selection and evaluation
are performed using ANOVA on Deviance (χ²), multicollinearity checks (VIF), and likelihoodratio tests, while model performance is assessed through ROC–AUC, confusion matrix,AIC/BIC,
and the Hosmer–Lemeshow goodnessoffit test.The results indicate that significant factors of
the outcome are related to navigation conditions and the context of the incident (for example,
weather/sea state, visibility, wind, event type/area, vessel type, number of pilots), based on the
interpretation of odds ratios and the statistical significance of the variables. Clustering analysis
indicated limited separability, suggesting the potential usefulness of alternative algorithms, such
as densitybased methods (DBSCAN/OPTICS), in future research. Overall, the study contributes
to a deeper understanding of the causes of maritime accidents and and helps strengthening their
prevention.
Postgraduate Studies Programme
Εφαρμοσμένη ΣτατιστικήDepartment
Σχολή Χρηματοοικονομικής και Στατιστικής. Τμήμα Στατιστικής και Ασφαλιστικής ΕπιστήμηςNumber of pages
76Language
GreekCollections
Except where otherwise noted, this item's license is described as
Αναφορά Δημιουργού-Όχι Παράγωγα Έργα 3.0 Ελλάδα
Αναφορά Δημιουργού-Όχι Παράγωγα Έργα 3.0 Ελλάδα