This work aims to develop a methodology for rotating machinery classification based on instance-based models. The development was based on data collected using a test bench that simulated four operating conditions. First, two models, linear regression and K-Nearest Neighbours (KNN) were chosen to perform the application to the case study, where linear regression was chosen to evaluate the difficulty of the problem. In addition, the feature extraction method used was principal component analysis (PCA). After the models were chosen, a hyperparameter optimization method with cross-validation (GridSearchCV) was used. Besides this, different percentages of variance explained were considered for the choice of the number of principal components. In a second step, a comparison of the linear and KNN model with literature results was performed through the optimization results. Finally, a new optimized KNN model was obtained, and, using metrics such as confusion matrix and accuracy, its resulting performance was competitive, considering efficiency and low complexity. The results show that KNN improved accuracy by 68 percent compared to the linear model.