Indoor occupancy detection and counting system based on boosting algorithm using different sensor data

This research examined how to estimate a room's occupancy utilizing datasets using data collected from various sensors with boosting methods. Different sensors were used to extract features that indicate the occupancy level and the relative changes (PIR motion detectors, CO2 sensors, plug loads, lighting loads, electricity use and Wi-Fi access points). The proposed two-layer method increases the reliability and accuracy of occupancy detection. Five machine learning models were trained and tested using data on occupancy (Logistic Regression, Multi-Layer Perceptron Classifier, Support Vector Machine, Light Gradient Boosting Machine and Gradient Boosting). Accuracy, Precision, F1-score, Recall and ROC were the performance parameters used. Thus, using the Python tool, the proposed optimized model gives better detection and estimated accuracy of occupancy in the indoor environment. As a result, data fusion in conjunction with the Light GBM algorithm's gradient boosting-based categorical features and PSO optimization techniques may be used to predict occupancy data with high accuracy and F1-score (both 99%).