Analysis of the mechanism of physical activity enhancing well-being among college students using artificial neural network

Abstract

This study explores the impact mechanism of college students' sports behavior on their well-being by constructing an Artificial Neural Network (ANN) model. The study employs an ANN architecture that combines a Long Short-Term Memory (LSTM) network and a Convolutional Neural Network (CNN). A prediction model is established based on the characteristics of sports behavior and psychological indices of well-being, such as psychological resilience, self-efficacy, and subjective well-being. The results show that the proposed LSTM + CNN model has achieved significant improvement on the test set. Its mean absolute error is only 0.072, the mean square error is 0.00596, and the root mean square error is 0.077, which is remarkably superior to traditional machine learning methods such as random forest and support vector regression. The innovative advantages of the proposed model in capturing the nonlinear relationships and deep characteristics of psychological and behavioral data is proved. The analysis of Shapley Additive Explanations (SHAP) values reveals three key factors significantly influencing well-being improvement. These impactful factors include the high-frequency exercise days per week (≥ 4), sustained morning exercise duration, and participation levels in group sports activities. The analysis of the dynamic threshold effect reveals that the critical points of distinct characteristic values exhibit substantial variations in their impact on well-being. Concurrently, the regulatory influence of sports behavior demonstrates differing intensities across diverse conditions. This study provides a new theoretical basis for designing personalized sports interventions and improves the accuracy of predicting psychological measurement data. Thus, it demonstrates the potential of sports behavior in promoting the mental health and well-being of college students.

Keywords: Artificial neural network; Psychological resilience; SHAP value analysis; Sports behavior; Well-being.

Fig. 1

ANN modeling between exercise and well-being.

Fig. 2

Pseudocode of the LSTM + CNN model.

Fig. 3

Descriptive statistics of sample information.

Fig. 4

Statistics of sample motion data.

Fig. 5

Descriptive statistical results of sample psychological test variables.

Fig. 6

Comparison between the true mean of psychological variables and the predicted mean of the model.

Fig. 7

Comparison of model accuracy evaluation indices.

Fig. 8

SHAP contribution degrees of sports behavior features to well-being indices (A: High-frequency exercise days per week (≥ 4); B: Cumulative duration of moderate-intensity exercise; C: Exercise heart rate variability, D: Proportion of nighttime exercise after 20:00; E: Daily step volatility, F: Morning exercise persistence cycle; G: Longest single duration of vigorous exercise; H: Exercise goal achievement rate; I: Frequency of recovery after exercise interruption; J: Density of exercise social interaction; K: Group exercise participation index; L: Exercise-sleep cycle synchronization; M: Daytime activity energy consumption gradient; N: Diversity of exercise scenarios; O: Decline rate of resting heart rate).