Accurate prediction of machine tool wear is an essential part of modern and efficient manufacturing. In recent years, many studies have been carried out using machine learning algorithms, both traditional and deep learning; with the latter ones reporting the highest precisions. The present work aims to show that, in the tool wear prediction problem, traditional methods can have a performance similar to the state of the art, obtained using deep learning methods. The data used here is presented in the form of time series, which cannot be used directly by traditional machine learning algorithms, such as the ones used in this work. To link the raw data and the learning algorithm, it is first necessary to extract a set of features from the time series. In addition, some preprocessing techniques, Bayesian hyperparameter optimization and forward feature selection are applied. In this work, two freely accessible databases are used with two different but related objectives, the first is used to predict machine tool wear, while the second is used to predict the remaining useful life of machine tools. For the first case, errors (RMSE) of less than 10 were obtained, while in the second case scores above 85% were achieved. In both cases, these results are comparable to the state of the art. Using the methodology presented here makes it possible to obtain very accurate tool wear predictions at a lower computational cost, both due to the use of less complex models and to a reduced set of features.