Comparison Analysis Of K-Nearest Neighbor (K-Nn) Algorithm With Naive Bayes For Fire Source Detection Mitigation

Abstract

Fire is one of the disasters that often occur in Indonesia. One of the consequences of fires that occur in Indonesia is forest fires. In 2014 and 2015 alone, 2.6 million ha of forest fires were reported in Indonesia. One way to detect a fire source is by developing machine learning that is used for information processing in the event of a fire by utilizing patterns or information from large data sets. This research will develop an algorithm to detect fires by comparing the accuracy of the two algorithms, that is K-Nearest Neighbor (K-NN) and Naive Bayes. The dataset was obtained from a fire simulation using NodeMCU ESP8266 and IR Flame Sensor, MQ7, and DHT 11. Based on the composition of the training and test data, this research found the best algorithm is K-Nearest Neighbor tuning using GridSearch CV, where the best metric parameters are ‘Minkowski’, K = 1, p = 1, and weights ‘Uniform’, with a composition of 75% training data and 25% test data with accuracy 96.44%, precision 96.48%, recall 96.44%, and F1-Score is 96.33%.