Comparison Performance of C4.5, Naïve Bayes and K-Nearest Neighbor in Determination Drug Rehabilitation

摘要

The popular data mining classification algorithm is C4. 5, Naïve Bayes and K-Nearest Neighbor (KNN). To be able to choose the best algorithm one of them can be measured from the level of accuracy, error rate and recall. This can be done by evaluation comparing several algorithms with one model of evaluation techniques. The data used for algorithm evaluation is taken from data of BNN drug rehabilitation of East Kalimantan Province with 550 data, which then divided into training data and data testing. The purpose of this research is to find the best algorithm in determining drug rehabilitation using a K-fold cross-validation algorithm with K = 10 for C4.5, Naïve Bayes and KNN algorithm. The research variables consist of medical status, employment status, work pattern, employment scale, duration of drug use, drug status, legal status, social status, psychiatric status, and rehabilitation status. Based on the results of the analysis by measuring the performance of the three algorithms using the confusion matrix method, with accuracy, error rate and recall, it is known that the best algorithm is Naïve Bayes with an accuracy percentage of 80.55% error rate of 19.45% and 70.10% recall.