| dc.description.abstract | Conventional home security systems often face limitations in automatically and accurately detecting threats in real time, making them vulnerable to crimes such as burglary, theft, and forced entry. This research develops an Internet of Things (IoT)-based home security system by integrating a vibration sensor (SW-420), sound sensor (GY-MAX4466), and magnetometer (GY-271), supported by the YOLOv8 visual object detection model and the Light Gradient Boosting Machine (LightGBM) classification algorithm. The system is designed to recognize suspicious activities around house doors, classify security conditions in real time, and deliver automated notifications to users through a web dashboard and Telegram application. Sensor data and YOLOv8 detection results are transmitted in real time to the Firebase Realtime Database and analyzed using LightGBM to classify incidents such as guest visits, person entry, burglary, theft, and attempted forced entry. The classification output is used as the basis for activating alarm buzzers and sending user notifications. System testing was conducted on three types of doors under critical scenarios, resulting in an accuracy of 91.25% for door 1 (180° swing) and perfect accuracy (100%) for doors 2 and 3 (90° swing). Overall, the LightGBM model achieved 97.09% accuracy with high precision and recall across almost all class categories. These results demonstrate that the proposed system is effective in enhancing home security automatically, adaptively, and is ready for deployment in real-world environments. | en_US |
