For residential, commercial, and industrial buildings to be safer, early detection of smoke and fire is essential. Conventional smoke detection systems frequently use straightforward threshold-based methods, which might result in false alarms or delayed responses. More sophisticated and dependable detection systems can be created with the development of Internet of Things (IoT) sensors and machine learning algorithms. By examining environmental sensor data, this study offers a machine learning-based smoke detection system intended to improve building safety. The suggested system makes use of several sensor characteristics that are gathered from IoT-enabled devices, such as temperature, humidity, gas concentration, pressure, and particulate matter levels. To detect the presence of smoke, a number of machine learning methods are used and assessed, including Random Forest, Gradient Boosting, AdaBoost, Logistic Regression, Support Vector Machine, Decision Tree, and K-Nearest Neighbour. A publicly accessible smoke detection dataset is used to train and evaluate the models. Results from experiments show that ensemble learning models—Random Forest in particular—perform better with high precision and recall levels. Additionally, a web-based tool is created that enables users to perform real-time smoke prediction, train models, and visualise datasets. The findings show that by making precise and data-driven decisions, the suggested method can greatly enhance early smoke detection and enable intelligent building safety systems.
Mahesh et al. (Sun,) studied this question.