Predictive Analysis and Segmentation of Child Malnutrition by Machine Learning in the Health Zones of Kasongo-Lunda (DRC)
Herman Mulasa Mbambu
Higher Pedagogical Institute Kasongo-Lunda (ISP - KASONGO-LUNDA), Kwango, Democratic Republic of the Congo.
Camile Likotelo Binene *
Department of Mathematics and Computer Science, Faculty of Science and Technology, National Pedagogical University, Kinshasa, Democratic Republic of the Congo.
Daniel Aluba Mulumba
Department of Mathematics and Computer Science, Faculty of Science and Technology, National Pedagogical University, Kinshasa, Democratic Republic of the Congo.
Guylit Kiala Lutumba
Department of Mathematics and Computer Science, Faculty of Science and Technology, National Pedagogical University, Kinshasa, Democratic Republic of the Congo.
Dephon Musele Busaki
Institute of Commerce of Idiofa. isc-idiofa (ISC – IDIOFA), Kwilu, Democratic Republic of the Congo.
Rostin Mabela Makengo
Department of Mathematics and Computer Science, Faculty of Science and Technology, University of Kinshasa, Kinshasa, Democratic Republic of the Congo.
Richard Kitondua Lubanzadio
Department of Mathematics and Computer Science, Faculty of Science and Technology, National Pedagogical University, Kinshasa, Democratic Republic of the Congo.
*Author to whom correspondence should be addressed.
Abstract
Child malnutrition remains a major public health challenge in the Democratic Republic of Congo (DRC), particularly in rural areas where access to adequate nutrition and quality healthcare remains limited. This study proposes a machine learning-based approach to model, segment, and predict the evolution of malnutrition in health zones of Kwango province, using data collected between 2015 and 2023. Four indicators were analysed: the number of moderately malnourished, severely malnourished, and severely malnourished children referred, and children aged five years and older. The K-Means classification algorithm identified two distinct groups: (i) a cluster with low prevalence and moderate progression, and (ii) a cluster with high prevalence dominated by the Kasongo Lunda area, representing the critical core of the phenomenon. The predictive phase, based on linear regression, made it possible to estimate trends up to 2026. The forecast results for the period 2024–2026 highlight an overall upward trend in child malnutrition across most health zones. Predictive modelling based on 2015–2023 trends reveals a near-linear progression of malnutrition in most areas. The projections indicate a sustained increase in malnutrition in several areas, with marked increases in Kasongo Lunda, Matamba Solo, Mahuangi and Muana Muyombo, while Kishiama and Mulundu show relatively stable levels. The areas show particularly alarming trajectories with an average annual growth rate of between 10% and 15%. The results demonstrate the relevance of machine learning techniques in nutritional surveillance. They offer a decision-making tool for planning health interventions, optimal resource allocation, and targeted prevention of nutritional crises at the community level. These results highlight the potential of data-driven health planning tools in supporting targeted, preventive, and resource-efficient nutritional interventions in rural areas of the DRC.
Keywords: Childhood malnutrition, machine learning, K-Means, linear regression, prediction, public health, health zones