Tribological properties such as friction and wear are still considered a problem in many engineering and scientific fields. While friction on the macroscale can be well described with appropriate equations and models, this is not the case for friction on the micro- and nanoscale. Given that the friction is a non-linear stochastic phenomenon that depends on several parameters, experimental measurements are more complex, as is the development of models that describe this phenomenon. The dependence of nanofriction on several parameters was thus studied in this work by using a scanning probe microscope. Experimental measurements of the friction force on the micro- and nanoscale were hence performed on two samples of thin films: aluminum oxide (Al2O3) and stainless steel (SS) in the experimental points defined by three variable parameters: normal force FN = 10 ... 2800 nN, sliding velocity v = 5 ... 3000 nm/s and temperature t = 20 ... 80 °C. Fifty measurement points were defined by employing the Centroidal Voronoi Tessellation (CVT) Design of Experiment (DoE) method in the multiscale space of variable parameters. From the experimental measurements of the friction force itself, it is not possible to determine its dependence on the three input parameters, and therefore the data obtained from the experimental measurements were analyzed by using different machine learning and artificial intelligence algorithms in order to obtain predictive models that connect the input parameters (sliding velocity, temperature and normal force) with the value of the friction force on the micro- and nanoscale. A statistical comparison of the developed machine learning and artificial intelligence models showed that of all machine learning models, linear regression has the best predictive performance, while of all artificial intelligence models, offspring selection has the best performances.