Hypertension remains one of the primary risk factors contributing to premature mortality on a global scale, affecting nearly half of adults, yet remaining undiagnosed in many cases. This paper studies the potential of machine learning and photoplethysmography (PPG) signals to classify and predict hypertension. Employing a dataset collected over a year at Guilin People’s Hospital [5], this research demonstrates a systematic approach to data preprocessing, feature selection, model training, and evaluation. Results indicate that combining machine learning algorithms with optimized signal preprocessing techniques achieves superior classification accuracy. Additionally, the integration of skewness-based Signal Quality Index (SQI) evaluation ensures high-quality signal selection, further enhancing model performance. The proposed method shows promise for scalable, non-invasive hypertension screening tools.