Recent developments in CAD/CAM (Computer Aided Design/Computer-Aided Manufacturing) systems have made it easy for unskilled workers to create NC programs. However, the determination of cutting conditions, which is essential for machining, still depends on the knowledge and experience of skilled workers. Therefore, this study aimed to discover tacit knowledge about cutting using data mining methods and to build a system to assist unskilled workers. Since cutting information is scarce because manufacturers have only recently commercialized their products, this study attempted to predict cutting conditions for lens tools by utilizing catalog information on barrel tools that have similar applications and designs. First, the databases of all tools were integrated by using a domain adversarial network. Second, new variables were introduced to eliminate domain dependencies. After confirming the validity of the new variables through repeated predictions, the common feature terms of barrels and lenses were considered. The results revealed domain invariance or domain dependence for each input feature, and common features across domains from the tool design and application of each tool.