Evidence record · UDINUS · SDGs
- Unit
- Rektorat UDINUS
- Period
- 2024
- Goals
- Record
- c3436208-9a32-4814-b5f9-5f5604e991c0
- Published
- 2026-10-30
Universitas Dian Nuswantoro, through researchers Ma'muriyah, Ni'matul; Noersasongko, Edi; Purwanto, Purwanto; Winarno, Sri; and Ashiddiq, Muhammad Ilham, conducted a study on data imputation and outlier detection for stunting classification. The activity addressed the importance of dataset quality in classification. The source explains that the Demographic Health Survey dataset is commonly used in stunting research but may contain missing and outlier data that negatively affect classification accuracy.
The study proposed applying imputation and outlier detection during the data pre-processing stage. Six imputation methods were considered: Max, Min, Mean, Mode, Median, and KNN Imputation. These methods were combined with three outlier detection methods: Boxplot, Z-score, and Interquartile Range. XGBoost was used for the classification task and was described as a machine learning algorithm known for its robustness in handling diverse data types.
Experimental results demonstrated that pre-processing through imputation and outlier detection improved stunting classification accuracy, achieving rates above 95%. The highest accuracy was obtained using median imputation. The activity therefore documents an analytical and computational approach to stunting classification through data preparation and machine learning. It provides evidence of research addressing stunting through methods focused on missing data, outlier data, pre-processing, and classification accuracy.