Abstract: Computational electromagnetics (CEM) occupies the core position in the field of electromagnetic engineering applications, such as optimization of radar target characteristics, analysis and design of antennas, and modeling of microwave devices and systems. Traditional CEM methods have matured after the researches over the past half century. In recent years, benefiting from the rapid development of artificial intelligence, intelligent electromagnetic computing has drawn an increasing attention as an emerging front. It aims to lower the technical implementation difficulty of traditional CEM methods, and greatly accelerate the simulation process under the condition of ensuring result accuracy. Focusing on the applications in electromagnetic scattering analyses, this paper discusses the intelligent electromagnetic computing methods based on the learning and extraction of target feature parameters, which are a group of parameters that connect the target structural attributes to the target electromagnetic characteristics. They can possess the functions similar to the system transfer function. Once obtained, they can be utilized to calculate the scattered far fields for arbitrary incident and observation directions. A brief outlook is given to this kind of methods in the near future.