Optimization of Population Document Services in Villages using Naive Bayes and k-NN Method

Abstract

The Indonesian government strives to improve its registration and the issuance of the population documents program. However, several obstacles are faced, such as the complicated topography of the area and the distance from the village. Therefore, ball pick-up services are urgently needed. The government of Alor Regency, East Nusa Tenggara Province, is one of the regions that has implemented this program. Nonetheless, due to constraints in both time and financial resources, not every village can served. Therefore, villages must be selected fairly so the program can run well. Machine learning, a classification technique using data mining concepts, is expected to address this problem. This research aims to identify the most effective method for classifying eligible villages. The experimental process includes preprocessing, model training using K-Nearest Neighbor (K-NN) and Naïve Bayes, and performance evaluation. The results show that both methods provide good results, albeit with slightly different levels of accuracy. Comparative analysis shows that K-NN has a higher accuracy rate of 97.14% for k=1 and k=2 on the Min-Max-normalized dataset but has the lowest accuracy of 77.1% at k=11 and k=13 on the raw dataset. In comparison, the NB method has an accuracy of 94.29% but is stable on raw and normalized datasets.