Preterm labor is a major challenge in maternal and neonatal health because it contributes to high rates of newborn morbidity and mortality. Early detection is crucial, but conventional static approaches often fail to identify risks accurately and in a timely manner. This study proposes the development of a dynamic machine learning-based preterm birth risk prediction model using the Long Short-Term Memory (LSTM) architecture combined with the Bayesian Updating approach. The model is designed to process multivariate time-series data from various clinical sources such as EHR (electronic medical record), EHG (electrohysterography), CTG (cardiotocography), and vital signals collected longitudinally during pregnancy. By leveraging LSTM's ability to capture long-term temporal relationships and Bayesian probabilistic renewal mechanisms, the model is able to provide real-time and adaptive estimates of preterm labor risk on a weekly basis. Risk prediction results are visualized in the form of interactive graphs with risk categorization (low, medium, high) to support fast and accurate clinical interpretation. The study used simulated data on 500 pregnant patients and showed that the system can adjust risk predictions as new data comes in. This research makes a significant contribution to the development of artificial intelligence-based clinical decision support systems for pregnancy monitoring. Going forward, integration with real clinical data and external validation in the hospital environment is expected to improve the accuracy and implementability of the system in daily medical practice.