Recent variations in the semiconductor device market have become increasingly intricate and are influenced by emerging artificial intelligence trends and national economic security policies. In semiconductor-related enterprise production management, the intercorrelations of device markets must be assessed dynamically in response to macroeconomic changes and novel applications. Time-varying multivariate vector analyses can address nonstationary dynamic market changes by extracting mutual coherency to evaluate dynamic correlations across segmented markets. However, standard coherence evaluation often fails to reveal authentic correlations and subtle structural relationships among segments because of common macroeconomic effects. This study proposes “frequency-selected time-varying coherency,” incorporating frequency selection based on a multiresolution analysis using the wavelet method, which demonstrates the feasibility of observing dynamic correlation structures among market segments.