Intelligent Detection of Office Occupancy Based on Hybrid Data Mining

This study aims to use hybrid data mining technology to realize office occupancy intelligent detection to improve space utilization, reduce energy consumption, and enhance employee efficiency. Firstly, K-means clustering is carried out on the original data after pre-processing by fusing images, sensors, and behavioral data, and the data is divided into different clusters. Secondly, the Deep Residual Attention Network (DRAN) is utilized to extract the feature from the data of each cluster. Finally, the Weighted Voting-based Extreme Learning Machine (WV-ELM) is employed to classify and detect the data, to realize the occupancy status identification with higher precision. The results show that the proposed model’s accuracy achieves 96.17%, at least an improvement of 6.23%. When iteration times are 100, precision, F1 score, and recall reach 91.30%, 90.96%, and 89.22%, respectively. Moreover, the proposed algorithm outperforms other comparison algorithms on Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE), among which RMSE decreases from 5.83 in 20 epochs to 4.66 in 100 epochs, MAE from 8.46 to 7.08, and MAPE from 10.52 to 8.00. Therefore, the detection system proposed in this study is superior to the existing technology in terms of accuracy, real-time, and robustness, and provides effective technical support for the space management and resource optimization of intelligent buildings.