BioAging-Opt

# Active Aging Intervention Optimization Framework This repository contains the code, datasets, and documentation for the research titled: **"Research on the Optimization of Sports Intervention Strategies for Active Aging and Identification of Mediating Effect Paths"**. --- ## Overview The project introduces a **novel computational framework** designed to optimize sports intervention strategies for active aging using advanced **biochip-based modeling** and **inference algorithms**. Key features of the framework include: - **Biochip Reaction Dynamics** modeling for biochemical simulation. - **BioGraphField** architecture for localized signal integration and spatio-temporal modeling. - **BioAware Inference Dynamics (BID)** for robust and interpretable causal analysis. This approach integrates **deep learning, causal inference, and explainable AI** to model complex nonlinear behaviors in aging populations, providing a foundation for designing **personalized and adaptive intervention strategies**. --- ## Key Highlights - **Explainable Causal Inference** Identification of mediating effect paths between physical activity and health outcomes. - **Adaptive Sports Strategy Design** Robust optimization across diverse aging profiles and social environments. - **Experimental Superiority** Outperforms state-of-the-art models such as ResNet50, SlowFast, VideoMAE, TimesFormer, and SwinVideo with improved **Accuracy**, **Recall**, **F1-score**, and **AUC** on multiple datasets (HRS, SHARE, NHANES, DAMMI). - **Modular and Extensible** Designed for reproducibility and future expansion, including advanced multimodal data inputs and real-time adaptation. --- ## Datasets The framework was evaluated on four benchmark datasets:1. **HRS** – Health and Retirement Study (USA). 2. **SHARE** – Survey of Health, Ageing, and Retirement in Europe. 3. **NHANES** – National Health and Nutrition Examination Survey. 4. **DAMMI** – Dataset of Activity Monitoring and Multimodal Interactions. These datasets include **biomarkers, wearable data, cognitive and social assessments**, and multimodal sensor streams. --- ## Repository Structure