Residential activity modelling has attracted considerable attention over the last years. This is particularly due to the fact that residential energy demand loads are highly dependent on the activity patterns of the household. Therefore, activity models are being increasingly used to underpin high-resolution energy demand models. This paper details the implementation of a new methodology for the analysis of empirical activity data that allows for the identification of characteristic behavioural patterns within them. The identified patterns are then used as the basis for the construction of a high-resolution residential user activity model. The model attempts to capture the statistical characteristics of the empirical data in the form of a stochastic process with memory of variable length. The proposed model is compared to a model based on the predominant first-order Markov chain approach. In addition to the modelling approach, a new metric for assessing the quality of activity sequences simulations is proposed. Given the amount of empirical data contained in any of the individual time-use datasets currently available, it would appear that the performance improvement over the predominant first-order Markov chain approach is modest. However, the validation results show that the proposed approach has the potential for broadening our understanding of the scheduling of activities in people's day-to-day lives and how this relates to the observed variability in both activity and energy consumption patterns.