Study of data-driven thermal sensation prediction model as a function of local body skin temperatures in a built environment

Personal comfort models – A new paradigm in thermal comfort for occupant-centric environmental control

“…To address the first two characteristics of personal comfort models noted above, the review only includes studies that focus on individual occupants as a unit of analysis, and use human feedback in the model development. This effectively excludes 1) studies that adopt a data-driven approach to modeling, but predict thermal comfort of general populations rather than individual occupants [18][19][20][21][22][23], and 2) studies that use synthetic data instead of real-world data to model individuals' thermal comfort [24][25][26]. Table 1 summarizes the findings from this literature review.…”

“…Hence, one might supplement surveys with objective methods of collecting individual-specific data to ensure consistency and quality of the data that can be integrated into personal comfort models. As examples, research shows that wearable sensors or connected devices can provide continuous data tracking of occupants' physiological conditions (e.g., skin temperature, heart rate) [45,18,46,21,47,23,27] or behavioral actions (e.g., personal fan use, thermostat adjustments) [25,27].  Challenging environmental measurement: Radiant temperature and air velocity are often omitted or simplified in the development of personal comfort models, largely because modelers intentionally target easily obtainable data and the instrumentation to collect these variables is costly.…”

“…Proxy variables that supplement or replace direct survey responses on thermal comfort after training are also a valid path. Research has shown correlations between individuals' thermal comfort survey responses and thermal control behavior (e.g., thermostat adjustments) [25] and physiological conditions (e.g., heart rate, skin temperature) [45,21,23,22,71], both of which can be measured continuously via non-intrusive monitoring technologies (e.g., smart thermostats, wearable sensors). Hence, they could potentially be used in personal comfort models as proxy variables to infer individuals' thermal comfort.…”

Personal thermal comfort models with wearable sensors

“…Second, most studies evaluated the performance of personal models, which predict categorical responses (e.g., cooler, warmer, no change), using accuracy that is the number of correctly predicted instances divided by the total number of instances in the dataset. Previous studies using such metric reported the prediction accuracy 79% ± 32% (Mean ± SD) for personal comfort models developed from physiological data with wearable sensors [18,19,29,32,39,41]. However, this metric is problematic because it fails to exclude correct prediction purely due to randomness [42,43].…”

Personal comfort models: Predicting individuals' thermal preference using occupant heating and cooling behavior and machine learning

“…With personally-owned thermal control devices such as PCS, we can trace the associated behavior back to individual occupants, creating a direct link to personal comfort. In addition to thermal control behavior, tracking occupants' physiological conditions via wearable sensors offers another convenient way of collecting additional data points about human thermal comfort, and recent studies [32][33][34] have used skin temperatures to predict individuals' thermal comfort. However, no studies have used records of occupant behavior with personally-owned thermal control devices for individuals' comfort predictions.…”