Investigation of well-motivated parameter space in Beyond the Standard Model (BSM) theories plays an important role in new physics discoveries. However, due to the curse of dimensionality, a large-scale exploration of models is typically a time-consuming and challenging task. In this project, we are refining an existing self-exploration parameter scan method, named Machine Learning Scan (MLS) [https://arxiv.org/abs/1708.06615] by incorporating uncertainty sampling and genetic algorithm. On top of that, we are also developing the method to be used over different problems via the principle of transfer learning. Such method are applicable for any scientific research that face problems with parameter scan.