Abstract
In recent years, with the country's emphasis on national fitness, the health status of primary and secondary school students has become the focus of social attention. As one of the important means to measure students' physical fitness, physical examination results are closely related to students' physical fitness. However, there are some problems in the traditional physical examination management, such as subjective influence, complicated manual calculation, and difficulty in retaining and making full use of data. Based on the physical fitness test data of primary schools in the past five years from 2018 to 2022, this study aims to apply machine learning and deep learning methods to deeply analyze and mine data information, provide automatic classification methods and accurate performance prediction models, and then expand to provide students with personalized training suggestions to assist teachers in making reasonable teaching plans and other applications. The first research method is the classification method based on BP neural network, which realizes automatic comprehensive grade classification and achieves 98.448% classification performance, and explores students' physical health and grade classification. The second research method is the performance prediction model based on CNN-LSTM neural network, which combines CNN feature matrix and LSTM continuous time series information to provide more accurate performance prediction for various physical test items, and provides a new method for the management and evaluation of physical test results of primary and secondary school students through data analysis and prediction model. These methods not only solve the problems of traditional evaluation methods, but also provide scientific guidance for schools and promote the healthy development of students and the optimization of physical education.