175.3 MB of experimental data, code, and results supporting the study 'LIME-DP: Instance-Adaptive Perturbations and Stability-Convergence Index for Reliable Gait Analysis Explanations.' The repository includes preprocessed datasets from HugaDB, GaitRec, and a primary IMU-based gait dataset, along with implementations and visualizations. Author Mercy Mulwa last updated the repository on April 23, 2026.
Use Cases
- Evaluating explanation stability and fidelity for gait analysis models based on the proposed Stability-Convergence Index (SCI)
- Comparing the performance of instance-adaptive perturbation frameworks (LIME-DP) against standard LIME methods
- Reproducing experimental results across Random Forest and MLP classifiers using the provided datasets and scripts
- Applying correlation-preserving perturbation techniques to improve local fidelity in clinical gait data explanations
Strengths
- Includes 175.3 MB of data, code, and visualization outputs supporting reproducibility
- Contains preprocessed datasets derived from three distinct sources: HugaDB, GaitRec, and a primary IMU-based gait dataset
- Provides a representative experimental script for generating explanations and evaluations
Limitations
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is unknown, which may limit suitability assessment
Provenance
- Source
- figshare
- Collection Method
- Preprocessed datasets derived from HugaDB, GaitRec, and a primary IMU-based gait dataset.
- Freshness
- Last updated 2026-04-23 11:32:30; freshness should be verified