Purpose <p>Diverse gait patterns in individuals with stroke require personalized rehabilitation to improve paretic propulsion (Pp), which reflects gait function. However, the gait characteristics associated with Pp have not been fully elucidated. This study aimed to identify gait characteristics in the paretic and non-paretic limbs associated with Pp, categorize gait patterns, and extract gait disease-specific characteristics related to Pp for each cluster.</p> Methods <p>Seventy-seven community-dwelling post-stroke patients and 101 healthy volunteers underwent three-dimensional gait analysis. Using Mahalanobis distance and Markov Chain Monte Carlo sampling, disease-specific gait characteristics associated with Pp were extracted and subsequently analyzed using multiple regression, path analysis, and k-means clustering. Gait characteristics related to Pp were identified in each cluster of the study.</p> Results <p>The paretic shank-to-vertical angle (SVA) at terminal stance, paretic maximum knee flexion angle, paretic ankle plantarflexion angle at toe-off, and non-paretic knee extension angle at initial contact were associated with Pp (<i>R</i>² = 0.744, <i>p</i> &lt; 0.001). Paretic SVA and ankle plantarflexion angle were significantly associated with Pp in cluster A with high gait function. Asymmetrical hip extension and non-paretic knee extension angles were significantly associated with Pp in cluster B, which had low gait function.</p> Conclusion <p>Paretic SVA and ankle plantarflexion angle may be associated with Pp and gait speed. The gait characteristics associated with Pp and gait speed may differ depending on ambulation patterns.</p>