Application of machine learning in sanitation management prediction: Approaches for achieving sustainable development goals

Abstract

Due to lack of understanding about the different conditions of sanitation systems and complexity, mostly in Asia and Africa, the occurrence of unsafely managed sanitation still exists, leading to severe environmental and health impacts. Because of the complexity of this problem, machine learning tools were applied to develop an effective model for health protection and safe sanitation management as a promising way to approach sustainable development goals. This study aimed to examine the incidences of ineffective sanitation management on the prevalence of diarrhea infections using machine learning tools. The prevailing conditions for safely sanitation management were identified, and the effective model to protect the impacts of ineffective sanitation management on the prevalence of diarrhea infections was proposed. Based on information collected from about 1000 households with relatively high diarrhea infections during the period of 2017–2021, factors relating to sanitation facilities for health protection and safe sanitation management were examined. Diarrhea infections and no diarrhea infection were recognized based on actual conditions of the surveyed households and incidences from ground-based observation. Classification tree model as J48 in WEKA was applied for analytic predictive tool using 70:30 ratio of training and validating dataset. The findings showed that the tree model obtained from the training data was with 73% accuracy prediction, while that from the validation data was with 70% accuracy. The correlation of personal hygiene (such as washing hand before meal and drinking water from natural water sources) and sanitation facilities (such as open defecation and distance from on-site sanitation facilities to open drains or storm sewer) was significant with inverse relationship for safely sanitation management and public health protection.