Abstract:

To improve the classification precision of rare class problem, by importing cost matrix and fitness function of GEP (Gene Expression Programming), we propose a  CSC-GEP(Cost-Sensitive Classification-Gene Expression Programming) . The experiments carried on benchmark datasets from the UCI machine learning repository show that CSC-GEP is effective. CSC-GEP increases the recall rate, precision and F-measure by 7.07%, 2.89%, and 5.12% respectively compared with C4.5.

Key words: classification, cost-sensitive, rare classes, gene expression programming