Increasing Smoke Classifier Accuracy using Naïve Bayes Method on Internet of Things

Abstract

This paper proposes fire alarm system by implementing Naïve Bayes Method for increasing smoke classifier accuracy on Internet of Things (IoT) environment. Fire disasters in the building of houses are a serious threat to the occupants of the house that have a hazard to the safety factor as well as causing material and non-material damages. In an effort to prevent the occurrence of fire disaster, fire alarm system that can serve as an early warning system are required. In this paper, fire alarm system that implementing Naïve Bayes classification has been impelemented. Naïve Bayes classification method is chosen because it has the modeling and good accuracy results in data training set. The system works by using sensor data that is processed and analyzed by applying Naïve Bayes classification to generate prediction value of fire threat level along with smoke source. The smoke source was divided into five types of smoke intended for the classification process. Some experiments have been done for concept proving. The results show the use of Naïve Bayes classification method on classification process has an accuracy rate range of 88% to 91%. This result could be acceptable for classification accuracy.