<p>Pregnancy termination (PT) is a sensitive and vital public health issue in developing countries. PT, whether spontaneous or induced, affects women’s reproductive health. This research aims to assess and compare various machine learning (ML) models for predicting and identifying factors associated with PT among reproductive-age women. This study used data from the 2022 Bangladesh Demographic Health Survey (BDHS). Five ML models were used: Extreme Gradient Boosting (XGB), k-nearest Neighbor (KNN), Support Vector Machine (SVM), Artificial Neural Networks (ANN), and Random Forest (RF). Of the total 8650 observations, 6920 (80.0%) were randomly selected for the training set, and the remaining 1730 (20.0%) were selected for the test set. To identify and select important features for analysis, the Boruta algorithm was used. Moreover, SHAP and LIME techniques were applied to identify features associated with PT, while statistical analysis was used to examine the distribution of PT. Among the five MLs, the random forest model (RFM) showed moderate performance, achieving the highest accuracy, specificity, precision, and AUC of 66.2%, 67.8%, 40.0%, and 71.8%, respectively. According to the RF model, significant contributing factors for PT include the number of children ever born, employment status, age at first sex, having a birth in the last three years, and the respondent’s age. The random forest model is expected to make a significant contribution to reproductive healthcare in Bangladesh. With the highest predictive performance, the ML tools help policymakers identify key predictors of pregnancy termination and formulate effective measures to reduce PT. This will, in turn, help implement the Family Planning Strategy 2025–2030, achieve SDG-3, reduce unintended pregnancies, and reduce the burden of pregnancy loss in Bangladesh and similar settings.</p>