Professional accreditation of engineering education has become a crucial tool for ensuring and enhancing the quality of engineering education. How to conduct scientific and effective evaluation and continuous improvement has become a prevailing topic. This study focuses on integrating the causal inference method into a graduation design evaluation system, adopting the causal diagram and the structural causal model as the core analytical tools, and selecting a case study of a university to reconstruct the teaching improvement process and assess the causal effect of the learning process on the quality of students’ graduation theses by gradually introducing the process evaluation. Meanwhile, through multivariate data analysis, the supervising instructor was identified as a confounding variable and this variable was deconfounded to reveal the true causal effects of each link. Continuous improvement measures were then proposed. The results demonstrated that the causal reconstruction process confirmed the instructional improvement measures from the data perspective, and the deconfounding method could effectively shield the effects of confounders. Through the application of the causal inference method, this study offers a multi-perspective and multi-level evaluation approach, overcoming the limitations of the traditional single evaluation method, and provides an empirical basis and strategic direction for the improvement of engineering education quality.