This project develops a predictive model to estimate room occupancy using environmental sensors. By leveraging sensors for CO2, temperature, sound, and light, the model provides real-time occupancy data, which can be used to drive intelligent building systems in both residential and commercial settings. This approach aims at enhancing comfort and optimizing energy efficiency.
The data for this project is sourced from the UCI Machine Learning Repository. It includes measurements from various sensors in a 6m x 4.6m room over a period of four days, collected every 30 seconds.
- Temperature: Readings from four sensors (Celsius)
- Light: Lux readings from four sensors
- Sound: Voltage readings from sound sensors
- CO2: PPM from a CO2 sensor
- PIR: Motion detection from two sensors
We explored several machine learning models:
- Naive Bayes: Utilizes a simple probabilistic model based on Bayes' theorem with strong feature independence assumptions.
- Support Vector Machine (SVM): Finds the optimal hyperplane for classification, used with gradient descent for optimization.
- Logistic Regression: Well-suited for binary classification problems, used here with a one-vs-all approach for multi-class classification.
The models' performance varied, with Logistic Regression showing the best overall accuracy and robustness across different classes.
Details on how to set up and run this project:
- Clone the repository:
git clone https://github.com/Omii2899/Room_Occupancy_Prediction.git
- Install the required packages
pip install -r requirements.txt