Feature Selection for the Shear Stress Classification of Hip Implant Surface Topographies

In order to determine the optimal hip implant surface topography that minimizes the shear stress during everyday activities, a large number of models need to be created and analyzed. To understand how different model parameters affect the shear stress values and distributions, the parameters are varied during the model creation process. Depending on the complexity of the model and the number of elements for the finite element simulation, the time needed to obtain the results can vary from a few minutes to a few hours. The aim of this study was to analyze the application of feature selection algorithms as a way to understand which parameters have the highest impact on the shear stress results. Two algorithms were considered – a tree-based model and Principal Component Analysis. The used dataset consisted of 64 models, previously analyzed using the finite element method. There were 11 input parameters and a single target variable. This approach can reduce the number of numerical models that need to be created and analyzed, thus saving time and resources. The obtained results indicated that the most important parameters can be extracted even when working with a small dataset such as the one considered in this paper.