<![CDATA[Village development has been one of the most important targets of government policies in Indonesia in order to fully optimize its potential. Under Law 06 Year 2014 on Villages, local governments from regency/municipality level to village level are required to understand their respective village potentials in order to increase the village potentials in their regions. In this paper, we build and analyze the Bayesian network methods to classify the village development index at regency/municipality and gain a better understanding of the causal relationships between independent variables of the village potential status. Using a web scraping method of information retrieval, data are collected from the Ministry of Village, Development of Disadvantaged Regions, and Transmigration (Kemendesa) website, and Village Development Evaluation (Indeks Pembangunan Desa—IPD) of Statistics Indonesia (BPS) publication in 2018 data. Further, we combine the discretization using the K-Means clustering method to handle the continuous nature of retrieved data. An extensive comparison of different learning structures of the Bayesian Network is performed, which includes the learning structure of Naive Bayes, Maximum Spanning Tree with weighted Spearman correlation coefficient, Hill Climbing search, and Tabu Search during the construction of Bayesian networks. For fairness evaluation, all constructed models are built using 80% data as a training set and the remaining 20% as a testing set. The results show that Bayesian network approach can be applied in village development index status classification where the construction using maximum spanning tree with K-Means data discretization gain the best performance of 90.69% accuracy.]]>
Copyrights © 2020
