The spectral dependence of the permeability of magnetic materials, whether in composite or dense materials, remains a complex issue due to the different scales of the phenomena involved. Approximate analytical models are often used to describe the frequency response of magnetic materials, particularly to improve their performance in areas such as power electronics. Recent results have shown that micro-magnetism codes can now predict the response of a system of coupled nanoparticles or a particle representing the volume of the materials in question. This thesis aims to use these tools to improve existing analytical models. An inclusion immersed in an effective field will be the paradigm from which the domain structure and the spectral response of the particle will be calculated using a micro-magnetism code. The materials studied include spherical particles or those with a high aspect ratio (magnetic oxides, ferromagnetic petals) at varying concentrations, ranging from dilute media to dense materials. This work will identify pathways to optimize the microstructure of materials for better performance in applications such as power electronics and microwave components. To this end, CEA provides a scientific computing environment with access to HPC resources, as well as facilities for sample preparation and static and dynamic magnetic characterization. At the end of this work, the candidate will have gained a solid understanding of the microstructure-property relationships described by a numerical approach applied to magnetic materials. More generally, this approach is expanding in the field of materials to improve their properties in various fields, under the designation "materials by design".