Occupancy pattern is one of the main factors influencing building energy consumption and is highly stochastic in nature. In the last few years,research has gained momentum to study and model the stochastic energy consumption pattern of occupants. Most of them discussed occupantswithin a specific space type (e.g., Office workplace) or activity type (e.g., opening/closing of a window or switching on/off lights). This paperpresents a study on the energy usage of different spaces within a building. Three institutional blocks are selected for this study with differenttypes of spaces such as class rooms, studios, computer rooms, offices, laboratories etc.). Space wise time series data of energy use (lighting andequipment loads) in kWh and occupancy density is collected for both semester and non-semester periods. A generic, feasible and low-cost Wi-Fi based method is used to collect the occupant density. A Classification and Regression Trees (C&RT) based model is used to establish therelationship between occupant density and energy usage at various spaces. Results of the preliminary analysis show significant variation inenergy usage pattern by the same group of people among different spaces. Multiple levels of occupancy modelling are proposed within a buildingto understand the interaction among different spaces. The results can aid the university facility management to make predictive and betterinformeddecisions for the operation and maintenance of the educational building spaces.