Model-Based Approach for Clustering Regencies/Cities in The Land of Papua Based on Food Security Indicators

The demand for food continues to increase as population growth concerns the Indonesian government, as stated in the second goal of the Sustainable Development Goals, namely zero hunger. The National Food Agency (BPN) uses the Food Security Index (IKP) to monitor food security conditions in Indonesia's district/city and provincial levels. Based on the BPN data, most districts/cities in The Land of Papua (so called Irian Province before the year 2000) are food insecure. However, the IKP has a weakness in the subjectivity of determining weights so that it can disguise the failure of a program or exaggerate a success. The model-based clustering (MBC) method can measure the food security of districts/cities in this area based on food security indicators. However, the data conditions are generally not multivariate distributed, and there are many outliers, so this study used MBC with multivariate t distribution because it is more robust. The best model was obtained with two clusters based on the largest Bayesian Information Criterion value. Cluster 1, located in the mountains and islands such as Nduga, Intan Jaya, Mamberamo Tengah, Puncak, and Lanny Jaya, had low food security, low indicator achievements with high poverty characteristics, many households with a portion of household expenditure on the food of more than 65%, low access to electricity and clean water, low life expectancy and average years of schooling for women, and the percentage of stunted toddlers. Meanwhile, Cluster 2, areas with high food security, had the opposite condition. Keywords: food security, model-based clustering, multivariate t distribution, Land of Papua