Induction hardening is a heat surface treatment technique widely employed for steel components in order to improve their fatigue life without affecting the metallurgy of the bulk material. The control of the treated components goes through the prediction and the optimization of the induction hardening process parameters. The aim of this work is to propose an approach based on artificial intelligence technique to predict the in-depth hardness profile. For this purpose, experimental tests were first carried out on 300M steel bar and C45 steel spur-gear under single and double frequencies, respectively. Intermediate variables were then generated to be used as input data. Data-driven model based on XGBoost library was finally developed. It was found that the proposed approach predicts with good agreement the hardness profiles and can be used in induction treatment process optimization.