<p>Scratch hardness has significant importance in investigating the tribological and wear properties of materials. This study aims to correlate scratch hardness with the plastic parameters of metals using empirical formulas and machine learning. 720 sets of finite element simulation of scratch tests considering the influence of interface friction coefficient were conducted. With the simulation results, empirical solution, Multi-Layer Perceptron neural network, and physics-informed neural network were developed for establishing the relationship between their scratch hardness and material parameters including dimensionless yield strength, strain hardening exponent, and the interface friction coefficient. Scratch and tensile tests of 18CrNiMo7-6, 304, 4140, 4340, DP600, DP800 alloy steel, and 6061 aluminum alloy were conducted to validate the prediction methods. The results demonstrate that the machine learning method outperforms the empirical solution in predicting scratch hardness. This study offers valuable insights for charactering scratch hardness of metallic materials, which has scientific and utilitarian significance.</p> Graphical abstract <p></p>