Household-level occupancy profiles generated using Bayesian Neural Networks for occupancy uncertainty in residential buildings

This dataset comprises time-series data related to occupancy availability and metabolic rates based on activities performed in residential dwellings, captured at 10-minute intervals. The data is structured into several CSV files, divided into two categories: metabolic rates and availability. The profiles are developed for two types of dwellings: house-type and apartment-type dwellings.
For house-type dwellings, there are 1000 CSV files each for metabolic rates and availability. These files are named sequentially from 1 to 1000. For apartment-type dwellings, there are 500 CSV files each for metabolic rates and availability, named sequentially from 1001 to 1500.
The metabolic rate files, such as DateTime_Metrate_1.csv to DateTime_Metrate_1000.csv for house-type dwellings and DateTime_Metrate_1001.csv to DateTime_Metrate_1500.csv for apartment-type dwellings, contain DateTime and metrate columns. The DateTime column records the timestamp at 10-minute intervals, and the metrate column records the metabolic rate in watts per person associated with the activity performed during the corresponding time interval.
The availability files, such as DateTime_Availability_1.csv to DateTime_Availability_1000.csv for house-type dwellings and DateTime_Availability_1001.csv to DateTime_Availability_1500.csv for apartment-type dwellings, contain DateTime and availability columns. The DateTime column records the timestamp at 10-minute intervals, and the availability column is a binary indicator where 1 denotes the individual is available (e.g., at home), and 0 denotes they are not available (e.g., outside).
The data originates from Time Use Survey (TUS) data, which records detailed activities of individuals over specified periods. Activities are classified based on their metabolic rates into Low Metabolic Activity (LA) for activities below 100 W/person, Medium Metabolic Activity (MA) for activities exceeding 150 W/person, NotActive for idle, resting, and sleeping states, and Outside Activity captured indirectly through the availability status.
This dataset supports research in occupancy prediction under uncertainty, energy consumption modelling in buildings, and smart home automation systems. To utilise the dataset, the metabolic rate and availability files should be integrated using the DateTime column. This alignment allows for a comprehensive view of both the activity level and presence status of individuals, which can be used to train and validate occupancy models.