Pattern recognition in non-weekly drop shipments at bpost Belgium An analysis based on time series similarity

Today bpost Belgium is confronted with deliveries of certain drop shipments at their deposit centers without announcement in advance on the date of delivery by the client. As a consequence, planning of the workforce to process the deliveries is difficult. The aim of this master thesis is therefore to detect patterns and/or trends in the non-weekly deliveries of drop shipments at bpost, in order to have a better insight into when deliveries will arrive and hence facilitate the workforce planning. The model developed to deduce the patterns is based on time series similarity exploiting four different distance measures. This format enables to find a data pattern with regular deliveries that closely matches the historical delivery pattern. The analysis is conducted on periodical level and aggregated on deposit center level in a second stage.