Comparing Machine Learning Models for Ukrainian Households Clustering through the Prism of Expenditures

Sammanfattning

Household expenditure patterns in Ukraine were explored with a focus on clustering techniques to uncover socioeconomic insights and inform policy interventions. Using an officially published Household Income Survey dataset for 2021, the research applies advanced machine learning models, including k-means, DBSCAN, and PCA, to analyze spending habits and identify household profiles with distinct expenditure priorities. The findings highlight significant variations in spending, such as elevated expenditures on food, housing, and transport, offering opportunities for targeted policies to redirect spending toward savings. The comparative analysis of clustering algorithms emphasizes their strengths and limitations, with k-means demonstrating simplicity and speed, while DBSCAN offers greater flexibility in handling complex data structures.