A hybrid feature selection approach for virtual metrology: Application to CMP process

This paper presents a feature selection method for virtual metrology applied to a chemical mechanical planarization process in the semiconductor industry. The proposed approach is based on a filter method coupled with a wrapper method. The goal of this article is twofold: (i) to improve the prediction accuracy of the considered machine learning algorithms, and (ii) to reduce the computational time and the storage space in databases. Numerical experiments are conducted on an industrial dataset derived from the chemical mechanical planarization process of a semiconductor manufacturing facility.