High-dimensional process data can make classification harder to manage. A correlation-based feature selection approach groups related attributes with Affinity Propagation and samples representatives, reducing feature counts while showing why validation remains essential.