In the wave of health technology, wearable sensors are attracting attention with their efficient health monitoring capabilities. This study uses big data technology to deeply analyze the data of these devices and mine the exercise behavior patterns. Through cleaning and preprocessing of large-scale datasets from multiple sources, combined with machine learning (e.g., random forest, support vector machine) and time series analysis, movement features and patterns are accurately extracted. Cluster analysis further identifies groups of exercise behaviors and provides insight into common and unusual patterns. Based on the personalized data, a movement habit adjustment model was developed to provide customized exercise advice to enhance exercise efficiency and reduce injury risk. Exercise compliance rates were generally high, with the majority of participants having compliance rates of more than 80%, suggesting that participants were able to follow personalized exercise recommendations well. This study demonstrates the potential application of big data and machine learning technologies in sports behavior analysis and health management.