Machine learning is a newly added tool in materials science that has the ability to significantly accelerate the discovery and design of materials with a large amount of data produced from experiments and high-throughput functional theory calculations. In recent years IKST researchers are focusing in studying in materials properties using machine learning based methods. Recently we have predicted the adsorption energies of mono-atomic and di-atomic gases on the surfaces of many transition metals (TMs) by using a machine learning approach. Our estimates of the adsorption energies are within within very root-mean-squared-error (RMSE) with less than 10 basic atomic features. Based on the important features of machine learning models, we have constructed a set of mathematical equations using a compressed sensing technique to calculate adsorption energy. Using similar approach we have predicted new III-V ternary semiconducting material with band gap close to the ideal value required for high efficient photo-voltaic applications.