The 15–24 age group is a key driver of Indonesia’s demographic bonus, yet the youth labor market continues to fluctuate due to limited skill readiness, economic conditions, and the absence of a strong labor force prediction system. These issues are compounded by sparse annual data from 2015–2024, which contains missing values and requires appropriate modeling to ensure accurate forecasting. This study aims to develop a prediction model for the number of young workers using three approaches—BSTS, LSTM, and SARIMA—while comparing their accuracy based on MAPE and RMSE. The study also tests whether performance differences among the models are statistically significant and identifies the most optimal model to support youth employment policy planning in Indonesia. The research uses time series data from BPS covering 2015–2024. Preprocessing includes imputing missing values through linear interpolation, normalizing the data, and dividing it into training and testing sets. Each model—BSTS, LSTM, and SARIMA—is then applied and evaluated using MAPE, RMSE, and statistical tests such as the Wilcoxon or paired t-test to assess the significance of performance differences. Results show that LSTM and SARIMA yield the highest accuracy, each achieving a MAPE of 3.44%, while BSTS performs less effectively on limited annual data. Statistical tests confirm that BSTS differs significantly from LSTM and SARIMA, whereas LSTM and SARIMA do not significantly differ from each other. The accuracy decline in 2020–2021 underscores the sensitivity of the youth labor market to external shocks like the pandemic. Overall, SARIMA and LSTM emerge as the most suitable models for forecasting youth labor numbers in Indonesia. These findings can guide the development of adaptive, data-driven employment policies to maximize the demographic bonus potential.