The stop-stop origin-destination (OD) matrix for a large multi-modal public transport (PT) network can be computationally challenging to process and analyse efficiently to support land-use and transport planning. We propose a novel hybrid approach that aggregates the PT stops by proximity and by the mobility pattern of each individual smartcard. The new stop groups improve the imputing of missing journey destinations and reduces the size and complexity of the PT OD matrix. It enables better activity location analysis and can be used for longitudinal cohort analysis to study the detailed changes of PT mobility pattern and behaviour across a network.