Objective: This work aimed to assess effects of different exercise intervention on Parkinson's disease (PD) treatment via a network meta-analysis. Methods: Eligible literatures were retrieved from three databases (PubMed, EMBASE and Cochrane Library) up to March 5, 2019 and screened based on established selection criteria. Afterwards, relevant data was extracted and heterogeneity tests were conducted to select appropriate effect models according to chi-square test and I2 statistics. Publication bias of included studies was also performed. Finally, the pairwise and network meta-analyses were carried out to evaluate the efficacy of different exercise training on PD management. Results: Overall, 19 studies encompassing 920 PD patients were identified to explore effects of interventions such as dance, Qigong, tango, resistance training (RT), Taichi and yoga on PD in terms of six indicators including six-minute walk, gait velocity, UPDRS III, PD questionnaire-39 (PDQ-39), timed up and go (TUG) and Berg balance test (BBT). The direct meta-analysis revealed that RT and dance altered the gait velocity and PDR-39 indicator of PD patients. And there was a statistical difference in RT and Tango regarding UPDRS III. Besides, significant differences were also detected among multiple comparisons based on TUG and BBT, containing RT vs control, Tai Chi vs control and Tango vs control for TUG, and dance vs control for BBT. Finally, results of network meta-analysis implied that tango was a good exercise for PD patients according to six different outcome measures. Conclusion: Tango was an optimal and effective option for improving functional mobility of PD patients. Keywords: Exercise interventions; Network meta-analysis; Parkinson's disease; Publication quality assessment; Tango.