Multi-sensor Data Fusion for Early Fire Estimation Using ML Techniques

Fire alarms are an essential aspect in providing safety for individuals and structures during a fire emergency. However, traditional fire alarms have several limitations, including false alarms and slow response times. In this study, we describe implementation and comparison of various techniques for machine learning like Naive Bayes, Random Forest, KNN, Logistic Regression, SVM Linear Kernel, and Decision trees to improve fire detection and response using various parameters such as Humidity, Temperature, MQ139, TVOC, and eCO2. There are many research papers which use deep learning and artificial intelligence using datasets containing images. However, our model has used a real-time-stamped text dataset and split them into train sets and test sets. Using various machine learning algorithms, different parameters have been calculated such as accuracy, precision, F1score, sensitivity, specificity, kappa, and RMSE values. Our findings suggest that fire alarms play an important role in smart home technology by providing early warning in the event of a fire and helping to protect people and property. Overall, fire alarms are becoming increasingly integrated into smart home technology, providing users with added convenience, safety, and peace of mind.