SportPsyCapNet

# Impact of Youth Sports and Psychological Capital on Long-Term Development This repository hosts the code, datasets, and experimental framework for the research paper titled: **"Research on the Long-Term Impact of Persistence of Youth Sports Participation Behavior and Accumulation of Psychological Capital on Individual Development"** ## 🧠 Overview This project introduces a novel computational framework that integrates psychological theory with state-of-the-art machine learning to examine how sustained youth sports participation shapes long-term individual development. By analyzing large-scale longitudinal datasets, we explore the interplay between physical activity and psychological capital (self-efficacy, resilience, optimism, and hope). ## 📌 Key Contributions - 🔧 **Hybrid Modeling Architecture**: Combines symbolic representations with deep learning (e.g., GRUs, transformers) to model behavioral trajectories.- 🧠 **Psychological Capital Integration**: Models psychological capital as an evolving state influenced by sports engagement.- 📊 **Multimodal Dataset Support**: Tested across ASDS, NLSCY, LifE, and MCS datasets with diverse formats and demographics.- 📈 **State-of-the-Art Results**: Outperforms existing baselines like SASRec, LightGCN, NeuMF on multiple metrics (accuracy, AUC, F1-score). ## 🗂 Datasets We utilize four prominent longitudinal datasets:- **ASDS**: Adolescent Sport and Development Study- **NLSCY**: National Longitudinal Survey of Children and Youth (Canada)- **LifE**: Long-term study on Education and Family (Germany)- **MCS**: Millennium Cohort Study (UK) ## 🏗 Architecture The core model includes:- **Contextual-Temporal Interaction Module (CTIM)**: Captures visual, temporal, and interactional dynamics in sports.- **Hierarchical Adaptive Knowledge Infusion (HAKI)**: Embeds semantic priors and temporal hierarchies.- **Graph Attention Networks (GAT)**: Models multi-agent interactions within sports environments.- **Uncertainty-Aware Refinement**: Enhances robustness under ambiguous or partial observations. See diagrams in the [PDF](./F1202-2-SR(1).pdf) for architectural visualizations (pages 9–14). ## 🔬 Experimental Setup - Framework: PyTorch with mixed precision and Weights & Biases tracking- Hardware: NVIDIA A100 GPU + 512 GB RAM- Optimization: AdamW, cosine decay scheduler, dropout, early stopping- Evaluation: Accuracy, Precision, Recall, F1-Score, AUC, MAE, RMSE ## 📊 Performance Our model achieved:- **ASDS**: 90.42% Accuracy- **NLSCY**: 91.13% Accuracy- **LifE**: 89.28% Accuracy- **MCS**: 90.37% Accuracy (Averages across 5 runs with low standard deviations) ## 🚀 Getting Started ```bash# Clone the repogit clone https://github.com/gxpGxdUb1amC/Impact-of-Youth-Sports.gitcd Impact-of-Youth-Sports # Install dependenciespip install -r requirements.txt # Run trainingpython train.py --dataset ASDS # Evaluatepython evaluate.py --dataset ASDS