Parkinson's disease (PD) is a prevalent neurodegenerative disorder with varying patient trajectories, yet little is understood about the underlying causes and symptom progression. The Parkinson's Progression Markers Initiative (PPMI) has collected comprehensive longitudinal data from diverse patient cohorts to identify biomarkers and aid in the development of interventions. Despite over 110 machine learning studies using the PPMI database, the majority have focused on supervised models for diagnosis prediction, which has limited impact on understanding patient variability and progression. This paper addresses this gap by combining supervised and unsupervised machine learning methods to identify subtypes that accurately predict disease progression in Parkinson's patients. Building upon previous work, we replicate and extend the study by integrating unsupervised patient clustering and prediction of present and future symptoms using 5 additional years of longitudinal data from the Progressive Parkinson's Markers Initiative (PPMI) database. Our findings demonstrate accurate prediction of disease trajectories and symptoms at baseline, offering valuable insights into patient heterogeneity and the potential for personalized interventions. The integration of supervised and unsupervised models presents a promising avenue for uncovering latent subgroups and understanding the complexity of Parkinson's disease progression.

本篇论文通过结合监督和无监督机器学习方法，识别出不同的亚型，能够准确预测帕金森患者的疾病进展和症状，并探究了病人异质性和个性化干预的潜力。

通过机器学习分析、识别和预测帕金森病亚型和进展