The worldwide aging population has been on the rise due to the advancement of medical science and decreasing birth rates. Unfortunately, this has led to an increase in falls among the elderly, which has emerged as one of the top ten causes of death for elderly individuals in Taiwan. To address this issue, our study aims to leverage the widespread use of smartphones and smartwatches to detect and prevent falls in older adults. Our machine learning model is trained using the UMAFall dataset, which collects data on daily activities and falling actions through five sensing devices. We used Azure Machine Learning Studio to test six different machine learning algorithms to evaluate suitable algorithms. In the end, the Two-Class Decision Tree algorithm achieved the best results. In our next step, this method will be programmed with the TensorFlow Lite and train the model. This model will be integrated into Android-based apps that can be installed on smartphones. By continuously connecting the smartwatch and smartphone Apps, the accelerometer, gyroscope, and magnetometer in both devices can be compared and analyzed in real time using the TensorFlow Lite model, accurately determining the wearer's actions.