In order to overcome phase separation problem of molten salt in long term service of the high-low temperature conversion, the eutectic salt mixture is usually suggested for heat transfer and energy storage media in concentrating solar power (CSP) plants. In the present work, the support vector machine (SVM) learning algorithm is firstly proposed to preliminarily predict the regularity of eutectic mixture formation of the AX-AY-AZ system, and then the eutectic properties of the NaCl-NaF-NaNO3 system are determined by thermodynamic modeling and thermal analysis method. On the basis of the SVM modeling, the NaCl-NaF-NaNO3 system is showed having a eutectic component and can form a eutectic mixture. In the meanwhile, the experimental results also indicate this salt mixture has a eutectic temperature and corresponding component, which confirms the accuracy of the predicted results by the SVM algorithm. Therefore, these theoretical modeling and differential scanning calorimeter (DSC) method can provides a preliminary screening test in the selection of molten salt thermal energy storage (TES) and heat transfer fluid (HTF) media in the CSP system.