Divorce is a global issue with profound emotional, psychological, and socioeconomic consequences. In 2022, Addis Ababa witnessed 14,000 registered marriages but also recorded 1,623 divorces, while 2018 saw 1,923 divorces. In Ethiopia, divorce prediction remain underexplored using machine learning approaches. Thus, understanding the factors contributing to divorce is vital for prevention and support. Machine learning and AI play a critical role in predicting divorce, early marital distress detection, and personalized interventions. This research explores a hybrid approach that combines AdaBoost, Gradient Boosting, Bagging, Stacking, XGBoost, and Random Forest with Jaya and Whale Optimization, along with an explainable approach. This hybridization allows for the benefits and strengths of each technique to be leveraged, leading to improved performance, interpretability, and insights into the underlying mechanisms of the divorce prediction. To assess model performance, traintest splits, and k-fold Cross-Validation techniques are used, with metrics like accuracy, precision, recall, F1 score, and AUC-ROC. AdaBoost achieves exceptional performance with an impressive accuracy, recall, and f1 score of 97.13% in both Jaya and WOA hyperparameter optimizations. Additionally, it attains high precision rates of 97.16% and 97.20% in Jaya and WOA optimizations, respectively. XGBoost outperformed the others in terms of AUC-ROC. Finally, We used SHAP and LIME explanations on Adaboost and XGBoost models to understand how it works and gain insights into its predictions. This research aligns with Sustainable Development Goals (SDGs) by promoting good health and wellbeing (SDG 3), identifying inequalities and offering targeted support (SDG 10), and fostering stable families and social cohesion (SDG 16). Keywords: Divorce, Ensemble Machine Learning, Nature-inspired Optimization Algorithms, XGBoost, LIME, SHAP, Sustainable Development Goals