Hybrid Clustering Model Using Fuzzy Logic and Machine Learning for Targeted Poverty Mapping

Accurately identifying impoverished households for targeted food assistance is a persistent challenge due to the ambiguous and multidimensional nature of socio-economic data. This study aims to address this challenge by proposing a hybrid clustering model that integrates fuzzy logic with multiple machine learning (ML) algorithms to improve the precision of poverty segmentation. The focus of the study is on leveraging fuzzy logic for preprocessing socio-economic data, transforming raw inputs such as income, education, health, and family size into degrees of membership to better reflect the continuous nature of poverty. The model incorporates K-Means, DBSCAN, and Hierarchical Clustering algorithms to perform clustering on fuzzified data. The results demonstrate significant improvements in clustering accuracy, with the highest Adjusted Rand Index (ARI) of 0.49 achieved using Hierarchical Clustering with five clusters, outperforming K-Means (0.28) and DBSCAN (0.09). This approach allows for a more nuanced differentiation between extreme poverty and moderate poverty, particularly across rural and urban settings. The study’s findings have practical implications for social policy, particularly in the context of sustainable development goals (SDGs) related to poverty alleviation. By enabling more accurate targeting of food assistance programs, the proposed hybrid model can contribute to more equitable resource distribution and inform data-driven policy decisions. This research underscores the importance of integrating fuzzy logic with machine learning for more effective poverty mapping and targeted welfare interventions.