Research on the Process Evaluation System of Higher Vocational Mathematics Driven by Big Data and AI - Guided by the Cultivation of Skilled Talents

Traditional assessment in higher vocational mathematics is often separated from the goal of skilled-talent cultivation. Summative assessment can hardly track students’ mathematical application ability in dynamic engineering-learning processes, nor can it diagnose skill gaps in time for targeted intervention. This limitation is particularly evident when mathematics courses need to support applied engineering tasks such as differential equations, signal modeling, electromagnetic-field computation, and antenna-related parameter analysis. To address this problem, this paper constructs a process-oriented assessment system driven by big data and artificial intelligence. First, multi-source data from students’ online learning behavior and offline classroom performance are collected and integrated through big data technology. Second, a long short-term memory network is used to dynamically model time-series learning behavior and identify weak knowledge points. Third, personalized learning paths are generated using knowledge graphs and collaborative filtering. Finally, a visual feedback mechanism transforms diagnostic results into teaching interventions, forming a closed-loop process of data collection, intelligent diagnosis, feedback, and adjustment. Experimental results show that the diagnostic accuracy of the LSTM model (RMSE = 1.66) is significantly better than that of baseline models. Personalized recommendation increases the improvement rate of weak knowledge points in the experimental group by 24.6%. The mathematical application ability score and professional project score of the experimental class reach 82.4 and 86.7, respectively, both significantly higher than those of the control class. The system realizes a shift from summative assessment to dynamic diagnosis and promotes the integration of mathematical ability with engineeringoriented professional skills.