USING DATA MINING TECHNIQUE TO PREDICT INFANT MORTALITY BY IDENTIFYING HIGH RISK BIRTH.docx

It has been widely recognized that one of the major cause of infant mortality is high-risk birth. Such can be identified through risk factors. Though, there are many risk factors, most interventions by government and agencies target birth based on a single risk factor (poverty) even though most infant deaths are not from the targeted group. Few studies have attempted to use multiple risk factors to predict infant mortality. Hence, data from Nigerian Demographic and Health Survey was gotten for this study and ten (10) risk factors were used to predict infant mortality risk by applying Random Forest algorithm to build a predictive model. A hybrid methodology of knowledge discovery process was used to achieve the required objective. Risk factors chosen for the study was compared with four (4) risk factors from a previous study. The result showed that targeting births with ten (10) risk factors gave an accuracy of 91.46%, sensitivity of 92.0%, area under curve of 96.1%, area under precision-recall curve of 94.6% compared to births targeted using four (4) risk factors having an accuracy of 88.6%, sensitivity of 83.6%, area under curve of 93.0% and area under precision-recall curve of 92.1%. The approach of this study gave an improved result by predicting more infant deaths compared to the previous study. This will aid governments and non-governmental agencies to target more births of high-risk most especially in Sub-Saharan Africa, where there are high needs but resources are low.