Mining Irregularities in Maritime Container Itineraries
Résumé
Identifying irregularities in sequential data is essential for many application domains. This paper discusses unusual events and how such events could be identified in sequential data. The type of sequential data used in this study holds location-based and time-based information. The irregularities are managed by establishing a weighted relationship between consecutive terms of the sequence. The sequences are spotted as irregular if a sequence is quasi-identical to a usual behavior which means if it is slightly different from a frequent behavior. This paper proposes a new approach for identifying and analyzing such irregularities in sequential data. The data used to validate the method represent cargo shipments. This work is part of a PhD research, now in the 3rd year. The proposed technique has been developed to identify irregular maritime container itineraries. The technique consists of two main parts. The first part is to establish the most frequent sequences of ports (regular itineraries). The second part identifies those itineraries that are slightly different to the regular itineraries using a distance-based method in order to classify a given itinerary as normal or suspicious. The distance is calculated using a method that combines quantitative and qualitative differences of the itineraries.