The ability to identify walking conditions correctly is essential for both diagnosing and treating gait abnormalities. This study utilized machine learning algorithms to analyze multivariate gait data obtained from 10 healthy participants walking under three distinct settings, including normal treadmill walking while wearing an ankle brace on the right. In order to increase the precision and efficiency of the ML models, the authors adopted a pipeline technique to handle the data. This research unveils the preeminence of random forest, achieving an impressive (92%) accuracy, surpassing logistic regression, neural network, naive Bayes, and perceptron. It exemplifies the formidable potential of machine learning algorithms for gait classification. The application of the results may be restricted to the particular dataset and walking conditions employed in the study, and the proposed study’s limitations include the use of a constrained number of algorithms and hyperparameter tuning settings. The proposed study has implications for the design of diagnostic tools and assistive devices for people with gait abnormalities and emphasizes the value of paying close attention to hyperparameters and other important model parameters to achieve the highest level of accuracy and performance in machine learning models. Future studies might build on this strategy by utilizing more datasets, additional algorithms, and sophisticated optimization methods to boost the precision of gait categorization.