Knowledge of absolute secondary electron yield (δ) is important for various applications of electron emission materials. Besides, it is also crucial to know the dependence of δ on primary electron energy Ep and material properties like atomic number Z. The available experimental database of δ reveals a large discrepancy among the measurement data, while the oversimplified semi-empirical theories of secondary electron emission can only present the general shape of the yield curve but not the absolute yield value. This limits not only the validation of a Monte Carlo model for theoretical simulations but also presents large uncertainties in the applications of different materials for various purposes. In applications, it is highly desirable to have the knowledge of the absolute yield of a material. Therefore, it is highly desirable to establish the relationship of the absolute yield with material and electron energy based on the available experimental data. Recently, machine learning (ML) methods have been increasingly used for the prediction of material properties mainly based on the atomistic calculations with the first-principles theory. We propose here the application of ML models to a material property study, starting with experimental observations and unfolding the relationship of δ with basic material properties and primary electron energy. Our ML models are able to predict δ(Ep)-curve covering a wide energy range of 10 eV-30 keV for unknown elements within the uncertainty range of the experimental data and can suggest more reliable data among the scattered experimental data.