Conceptual Modeling for Semantically-Guided Feature Selection in Machine Learning: A Case Study on Mohs Micrographic Surgery

The problem it addresses

The use of Machine Learning techniques in healthcare is transforming clinical practice by turning large volumes of data into predictive and diagnostic tools. However, their application depends heavily on data quality and availability. One of the main challenges is the curse of dimensionality, which arises when the number of features is large relative to the number of available samples. Feature selection techniques are commonly used to mitigate this issue, but many overlook the semantic relationships between variables. This work addresses this limitation through conceptual modeling techniques that organize clinical variables into semantically coherent groups to be used as inputs for Machine Learning models.

My contribution

Contributed to the development of the conceptual model and the application of a semantically-guided feature selection approach, supporting clinical domain representation and the organization of biomedical variables into coherent groups. The approach was applied to a Mohs micrographic surgery use case in collaboration with the Valencia Oncology Institute (IVO) to define the input features for predictive models.

Main results

The work demonstrates how conceptual modeling can incorporate clinical knowledge into Machine Learning experimental design. The case study shows how semantic feature grouping helps define clinically meaningful sets of variables for predictive models, complementing traditional statistical feature selection methods.