Ant Colony Optimization for Clustering College Students’ Physical Exercise Behavior Patterns

Abstract

This study employs a multi-channel data collection strategy, including questionnaire surveys, intelligent bracelet data, and interview observations, to comprehensively understand the characteristics of college students’ physical exercise behavior. By analyzing 2000 valid questionnaires and smart bracelet data from 500 students, we identified distinct patterns in exercise preferences and behaviors. Specifically, male college students tend to favor high-intensity exercises and team sports, whereas female students generally prefer moderate-intensity activities such as yoga. Furthermore, the proportion of high-intensity exercise initially increases and then decreases as students progress from freshmen to seniors, while the proportion of low-frequency exercise steadily rises, suggesting the impact of increasing academic pressures and changing life rhythms on physical activity levels. To enhance the analysis, we utilized an Ant Colony Optimization (ACO) algorithm to process the collected data. The ACO algorithm achieved a Silhouette Score of 0.70 and a Davies-Bouldin Index of 0.75, indicating superior clustering quality compared to traditional methods such as K-Means (Silhouette Score: 0.62, Davies-Bouldin Index: 1.0) and DBSCAN (Silhouette Score: 0.68, Davies-Bouldin Index: 0.8). Through this advanced clustering technique, we identified five distinct clusters with clearly defined features, enabling a more accurate identification and description of different exercise behavior patterns. These findings highlight the effectiveness of our ACO-based clustering method in capturing nuanced differences in exercise behaviors and provide insights into the varying preferences and routines of college students across different demographic and academic stages.