Digital transformation is a high-risk, long-term and systematic organizational change, which is highly dependent on the cognition and behavior of strategic decision-makers. Based on the Upper Echelons Theory, this research uses the machine learning ensemble algorithm to take China’s A-share listed companies from 2015 to 2022 as the research object, comprehensively investigates and compares the predictive ability of CEOs’ multi-dimensional personal characteristics on the level of enterprise digital transformation from 5 dimensions and 16 elements, and it identifies the key personal characteristics of qualified “navigators” of digital transformation. The results show that: (1) CEOs’ personal characteristics can predict the level of enterprise digital transformation, and the prediction ability of different characteristics is different. (2) From the comparison between feature groups, in terms of digital technology and digital output level, characteristics of CEO ability have the highest predictive ability; In terms of digital applications, CEO psychological characteristics have the highest predictive power. (3) From the perspective of individual characteristics, the five characteristics of CEO ownership, openness, tenure, overconfidence, and age show the best prediction effect, which are the key personal characteristics to promote the improvement of the digital transformation level of enterprises.