<p>In response to the problem of restricting team members' individuality and comprehensive development in the discipline competition teaching mode, this paper proposes a Competition Team Member Category Prediction (CTMCP) model based on the Multi-strategy Improved Coati Optimization Algorithm (MICOA) optimized Stacking model. For COA, firstly, the cyclone foraging strategy is introduced to expand the population search range; secondly, a defense mechanism against predators is proposed to enhance the ability to balance exploration and exploitation; then, the soft-frost search strategy is adopted to accelerate the algorithm convergence speed; finally, Cauchy mutation is applied to the optimal individual to avoid falling into local optima. The results of multiple sets of simulation experiments on the CEC2017 test functions comparing MICOA with other algorithms demonstrate that MICOA exhibits stronger optimization performance, providing strong support for the hyperparameter tuning of MICOA in machine learning single models and Stacking models. For the research on predicting the categories of competition team members, firstly, factor analysis is used to reduce the dimension of various data of competition team members and define relevant factors, secondly, the clustering result of factors is used as the label for constructing the prediction model of competition team members, then, the original data of the first 3 and first 5 semesters of team members are selected as features, finally, based on the MICOA-optimized machine learning single model, the prediction of team member categories achieves a significant improvement in classification performance. The prediction based on the MICOA-optimized Stacking model further improves the classification performance, validating the effectiveness of the proposed model in predicting the categories of competition team members.</p>