Figure 20
The process begins with “1. Simulation Values,” consisting of five physical environmental inputs: “Temperature (Celsius),” “Relative Humidity (Percentage),” “C O 2,” “Illuminance (Lux),” and “Sound Level (Decibels).” These feed into “2. Regression Values,” which include seven subjective or psychological factors: “Thermal Sensation,” “Thermal Satisfaction,” “Air Quality Perception,” “Daylight Perception,” “Daylight Satisfaction,” “Sound Perception,” and “Sound Satisfaction.” Arrows from “Temperature” point to “Thermal Sensation” and “Thermal Satisfaction.” Arrows from “Relative Humidity” point to “Thermal Sensation,” “Thermal Satisfaction,” and “Air Quality Perception.” An arrow from “C O 2” points to “Air Quality Perception.” Arrows from “Illuminance” point to “Daylight Perception” and “Daylight Satisfaction.” Arrows from “Sound Level” point to “Sound Perception” and “Sound Satisfaction.” All the factors under “Regression values” serve as inputs to “3. M L Prediction,” represented by a neural network model with distinct input, hidden, and output layers. The model has seven input nodes on the left, two hidden layers with multiple nodes in the middle, and an output layer with three nodes. All layers are fully interconnected. Finally, the output of the “M L Prediction” stage leads to “4. Score,” representing the final “Comfort Dimension Satisfaction” metric, displayed on a scale from 0 to 2.Diagram illustrating the application of the ML model within the tool prototype. Source: Authors' own work