Peer-reviewed article
Clustering Indonesian Provinces on Prevalence of Stunting Toddlers Using Agglomerative Hierarchical Clustering
Septian Wulandari ·
Summary
Written by AI, not yet AICS's view · claude-haiku-4-5-20251001This study uses statistical clustering methods to group Indonesian provinces based on characteristics related to child stunting, malnutrition, obesity, and immunization rates. Data from 2018-2022 across 34 provinces were analyzed using hierarchical clustering algorithms. The analysis identified two clusters: 13 provinces with high stunting prevalence and 21 provinces with low stunting prevalence. The study aims to help the Indonesian government target prevention and intervention efforts more effectively.
Key points
- Indonesia has the 5th highest number of stunted toddlers globally, with 30.8% prevalence in 2018
- The analysis grouped 34 Indonesian provinces into two clusters based on stunting, malnutrition, obesity, and immunization coverage
- Ward's algorithm clustering method achieved a cophenetic correlation of 0.8399978, outperforming the Average Linkage method
- 13 provinces were classified as having relatively high stunting prevalence
- 21 provinces were classified as having relatively low stunting prevalence
Limitations
This is a descriptive clustering study without intervention evaluation, so it cannot demonstrate what causes stunting or what solutions would be effective. The document does not provide details on specific provincial characteristics, the 11 data attributes used, or recommendations for action based on the clustering results.
Abstract
Stunting is a chronic nutritional problem caused by a lack of nutritional intake in toddlers. Indonesia is the 5th country with the highest cases of toddler nutrition experiencing stunting at 30.8% in 2018. The current problem, in Indonesia, is providing complete immunization and fulfilling child nutrition in each province is still low. Data obtained from 2018 to 2022 still toddlers who are malnourished and obese and there is no province grouping based on characteristics such as malnutrition, obesity, short toddlers, and complete basic immunization. Clustering is grouping objects into a group so that one cluster contains objects that are similar and different from other objects in other clusters. The agglomerative hierarchical clustering method can classify provinces based on the characteristics that cause stunting so that it can be used as a basis for early prevention for the Indonesian government to tackle stunting and can reduce stunting growth rates which continue to increase and can experience a decline. The agglomerative hierarchical clustering method used is the Average Linkage and Ward's algorithms with the data used is the prevalence of stunting taken in 34 provinces in Indonesia with 11 data attributes. The results of this study are that there are two clusters, namely Cluster 1 which has a relatively high prevalence of stunting with members of 13 provinces, and Cluster 2 which has a relatively low prevalence of stunting with members of 21 provinces. The highest cophenetic correlation value is in Ward's algorithm with a value of 0.8399978. So, it can be said that Ward's algorithm is better than the Average Linkage algorithm in clustering provinces in Indonesia on the prevalence of stunting toddlers.