norrth-z/UA-IMVC: Code and Data
deep Learning.py contains the core algorithmic implementation of the proposed Uncertainty-Aware Intelligent Multi-View Clustering (UA-IMVC) framework. The script integrates view-specific deep autoencoders to extract low-dimensional nonlinear latent representations from heterogeneous clinical features, coupled with an evidential deep learning module that estimates Normal-Inverse-Gamma (NIG) parameters to explicitly quantify both aleatoric and epistemic uncertainties. Furthermore, it incorporates a reliability-guided adaptive feature fusion mechanism to aggregate view-specific insights based on their normalized inverse uncertainties, a confiden
Resource information
- Format
- zip
- Original repository
- 20070811
- DOI
- 10.5281/zenodo.20070811
- Credit basis
- Creators in Zenodo record metadata
- Access
- Open files available from the original Zenodo deposit
- Last checked
- Nov 5, 2026
- License
- cc-by-4.0
- Last record update
- Nov 5, 2026
About this resource
deep Learning.py contains the core algorithmic implementation of the proposed Uncertainty-Aware Intelligent Multi-View Clustering (UA-IMVC) framework. The script integrates view-specific deep autoencoders to extract low-dimensional nonlinear latent representations from heterogeneous clinical features, coupled with an evidential deep learning module that estimates Normal-Inverse-Gamma (NIG) parameters to explicitly quantify both aleatoric and epistemic uncertainties. Furthermore, it incorporates a reliability-guided adaptive feature fusion mechanism to aggregate view-specific insights based on their normalized inverse uncertainties, a confidence-driven ensemble clustering strategy to achieve stable and reproducible patient stratification, and a post-hoc SHAP (SHapley Additive exPlanations) interpretability module via a LightGBM surrogate model to decode the "black-box" model predictions into transparent clinical logic. The file swelling.xlsx contains a literature-derived dataset developed to investigate observational indicators relevant to wound swelling through uncertainty-aware multi-view clustering. The dataset was constructed by systematically reviewing 227 medical articles and extracting 547 distinct observational indicators. These indicators were organized into five complementary semantic views: pain, function, morphology, biochemical, and other. Representative indicators include the Visual Analogue Scale (VAS), tumor necrosis factor (TNF), interleukin-6 (IL-6), swelling, and knee range of motion (ROM). The dataset is represented by a binary literature–indicator co-occurrence matrix, in which each row corresponds to a reviewed article, each column represents an observational indicator, and each entry indicates whether the corresponding indicator was reported in that article (1 = reported; 0 = not reported). Indicator names were harmonized by consolidating synonymous terms and spelling variants. The resulting literature-derived data provide the basis for uncertainty-aware multi-view representation learning, clustering, and the subsequent identification of indicators that distinguish the discovered literature clusters. This dataset captures indicator-reporting patterns in the reviewed literature rather than patient-level clinical observations.