During the research and development (R&D) stage of semiconductor fabrication, the R&D engineers make a lot of effort to identify golden dice that meet simulated performance of circuit design. With feedback from wafer acceptance test (WAT) data of the golden dice, the efficiency of process window analysis can be further improved. However, it is difficult for current practices to select golden dice due to limited time and cost concerns. In this research, an analytical model is proposed to analyze WAT data during the R&D stage of semiconductor fabrication to assist R&D engineers in resolving these critical issues. WAT data are collected and utilized to classify dice on a wafer and similar golden dice are then identified based on pre-defined golden dice. Similar golden dice can provide much more important feedback from WAT data, and the efficiency of process window analysis can then be improved. Real WAT data at the R&D stage during semiconductor fabrication were collected from a famous semiconductor manufacturing company and were experimented through the presented methodology. Experimental results show that the presented model can successfully extract representative similar golden dice within clusters. With advice from R&D engineers, the representative similar golden dice extracted from this work are sufficient for subsequent process window analysis.