Machine learning is getting more and more used nowadays, also in mechanical engineering. This thesis presents algorithm of research publications text classification using machine learning. It is shown the preparation of text step by step, which is very important before the actual classification because of the later performance of the algorithm. Texts used in this algorithm are research publications, which co-authors are employees in Faculty of mechanical engineering, University of Ljubljana. We classified the publications into ten different classes from the international academy for production engineering – CIRP. The most common used categories and most used words are shown and presented.