Predicting Unintended Pregnancy in Senegal by using Machine Learning Models: Evidence from Senegal DHS 2023

Unintended pregnancy refers to a pregnancy that is either mistimed or unwanted at the time of conception. Such pregnancies can have harmful effects, including negative impacts on maternal and child health, economic hardship, and strained relationships. This study aims to assess the effectiveness of machine learning algorithms in forecasting unintended pregnancies in Senegal and identifying the key factors that significantly impact them. The study utilized data from the 2023 Senegal Demographic and Health Survey, focusing on pregnant women. A final sample of 885 respondents was analyzed after handling missing values and six machine learning models namely Logistic Regression, Random Forest, K-Nearest Neighbors, Support Vector Machine, Naive Bayes and Extra Tree Classifier was used. The Random Forest algorithm emerged as the best predictive model due to its highest AUC value (80.45%), surpassing all other machine learning algorithms used in this study. Total birth, currently residing with husband/partner, respondents education level, number of living children, Husband/partners occupation, residence type, Respondent can refuse sex and intention of contraceptive use are identified as contributing factors in that predict unintended pregnancy. The studys results suggest that machine learning models, particularly Random Forest, can significantly enhance predictive accuracy for unintended pregnancies, helping public health initiatives in Senegal target at-risk populations. By identifying women at risk of unplanned pregnancies, targeted interventions and support services can be implemented, ultimately improving maternal health outcomes.