Predicting Contraceptive Usage for Married African Women Residing in Rural Areas: A Comparative Study of Deep Learning and Machine Learning Models with XAI Insights

Abstract

Abstract Background The use of modern contraceptives is a crucial aspect of family planning, especially for women residing in rural areas of Africa. This study seeks to address the issue of contraceptive usage among married African women living in rural areas by exploring the effectiveness of machine learning and deep learning models for predicting this usage. Methodology The data used in the study was obtained from the Multiple Indicator Cluster Survey 6 (MICS6) to develop and compare machine learning and deep learning models for predicting contraceptive usage among married African women residing in rural areas. In addition to predictive accuracy, the study also focused on incorporating explainable Artificial Intelligence (XAI) insights to enhance the interpretability of the models. Result Artificial Neural Network (ANN) with BFloat16 and Convolutional Neural Network (CNN) + Long Short-Term Memory (LSTM) were the best predictive models with 75% accuracy in both models and F1-score of 73% and 74% respectively. Additionally, XAI techniques provide valuable insights into the factors that influence contraceptive usage in the target population. Conclusion The results of this study indicate that deep learning models outperform traditional machine learning models in predicting contraceptive usage among married African women residing in rural areas. The implications of this research are significant, as the findings could inform policy and intervention strategies aimed at improving family planning services in rural areas of Africa.