This project aims to model and predict large-scale population dynamics in Sweden, including marriage, divorce and childbirth, using the Swedish Population Registry Data from 2015 to 2023. The primary task involves predicting which individuals will get married at each year, based on the previous years data. We train XGBoost models using historical data and construct pairwise features that capture both individual (e.g. age, education) and relational characteristics (eg. age differences, income gap). Additional models are developed to predict divorce risks and childbirth timing and frequency among couples. The resulting predictive models will be finally integrated into our existing individual-based simulation framework, which was previously developed to study the long-term health impacts of a digital alcohol intervention. By adding such population dynamics to the simulator, we can model intergenerational health effects such as how parental alcohol use or lifestyle behaviors may influence child outcomes over time.