A Prediction Model of College Students' National Identity was Constructed and Compared Based on Machine Learning

Objective: To explore the influencing factors of college students' ethnic identity, screen the relevant variables that affect college students' ethnic identity, establish a prediction model for college students' ethnic identity based on three algorithms, and compare the predictive value of these three models. Method: A questionnaire survey was conducted on college students from six universities in different regions of Jilin Province, with reference to Phinney's National Identity Scale, to collect relevant information on students' sense of ethnic identity. Using Lasso regression to screen the variables most related to college students' ethnic identity, prediction models for college students' ethnic identity were established based on support vector machine, logistic regression, and decision tree algorithms. Draw ROC curve and compare the discrimination of prediction models. Result: The Lasso regression results showed that 18 variables were most correlated with college students' ethnic identity. The analysis results of the risk model construction showed that the AUC of logistic regression, decision tree, and support vector machine for predicting college students' ethnic identity were 0.76, 0.666, and 0.848, respectively. Conclusion: Classroom education, participation in practical activities, and promotion of campus culture are important dimensions that affect college students' sense of national identity; The support vector machine algorithm has the best discriminative power in constructing a prediction model for college students' ethnic identity.