The paper presents a comparative analysis of selected algorithms for prediction and data analysis. The research was based on data taken from a computerized numerical control (CNC) milling machine. Methods of knowledge extraction from very large datasets, characteristics of classical analytical methods used in datasets and knowledge discovery in database (KDD) processes were also described. The aim of the study is a comparative analysis of selected algorithms for prediction and data analysis to determine the time and degree of tool usage in order to react early enough and avoid unwanted incidents affecting production effectiveness. The research was based on K-nearest neighbor, decision tree and linear regression algorithms. The influence of the rate of learning and testing set sizes were evaluated, which may have an important impact on the optimization of the time and quality of computation. It was shown that precision decreases with the increase of the K value of the average group, while the percentage of the number of classes in a given set (recall) increases. The harmonic mean for the group mean also increases with increasing K, while a significant decrease in these values was observed for the standard deviations of the group. The numerical value of accuracy decreases with increasing K.