A Study and Analysis of Predicting College Students’ Final Exam Scores by Integrating Physical Fitness Test Data and Poverty Level Information

To accurately predict the final exam scores of college students, a prediction model based on random forest is constructed by collecting data that may affect the final exam scores from multiplatform databases. The performance of this model is compared with the traditional neural networks and support vector machine model. At the same time, to further optimize the model, a feature selection method based on the random forest is proposed. The number of input parameters was increased or decreased in turn to establish a new model. The contribution of each input parameter to the overall model was calculated through the changes in the accuracy of the models before and after. The parameters required for modelling are selected according to the contribution of each parameter; in this manner, the model is simplified. The results show that the random forest model is better in fitting ability, generalization and fitting accuracy than the traditional neural networks and support vector machine. Feature selection can effectively simplify the model and improve prediction accuracy. Through the verification on the data of students in another grade, the model is universal.