An Artificial Neural Network Model for Predicting Children at Risk of Defaulting from Routine Immunization in Nigeria

Abstract

It has been widely recognized that immunization remains one of the most successful for decreasing child mortality rates and preventing several serious childhood diseases globally. This study proposed a prediction model for accurate identification of routine immunization defaulters in Nigeria. The proposed framework classified defaulters at five different risk stages: insignificant risk, minor risk, moderate risk, major risk and severe risk to reinforce targeted interventions by accurately predicting children at risk of defaulting from the immunization schedule. Data from Nigerian Demographic and Health Survey 2018 was obtained for this study and thirty-four (34) demographic and socio-economic factors were used to predict children at risk of defaulting from routine immunization in Nigeria by using Artificial Neural Network (ANN) to train the dataset. The results indicated that ANN model produced an accuracy of 99.16% for correctly identifying children who are likely to default from immunization series at different risk stages. Other performance measures include Precision of 99%, Recall of 99% and F1 Score of 99%. The model was further validated using one thousand (1000) dataset, out of which nine hundred and seventy four (974) were correctly predicted.

Keywords

• artificial neural network (ann), immunization, vaccination, immunization defaulters.