Simulation data for behavior-enhanced network clustering and data-driven robust consensus in large-scale group decision making
Description
This dataset provides numerical simulation inputs and archived results for behavior-enhanced network clustering and data-driven robust consensus in large-scale group decision making. It supports the numerical example, sensitivity analyses, comparative analyses, component ablations, and supplementary validation reported in the accompanying manuscript. The data include initial expert evaluations, incomplete directed trust information, behavioral parameters, attribute weights, adjustment-cost samples, clustering outputs, consensus trajectories, and optimization results. Extended validation covers attribute-weight perturbations, out-of-sample cost evaluation, scale experiments with different group sizes, and a pure-similarity network configuration. Run records retain parameter settings, random seeds, solver information, and termination outcomes. Files are provided as NumPy arrays (.npy) and JSON records, accompanied by an English README and a SHA-256 checksum manifest. The documentation explains variable definitions, indexing conventions, cost scaling, measurement conventions, and correspondence with the manuscript and supplementary material. All data originate from numerical simulations. This deposit retains the original 17 data files available at https://doi.org/10.17632/fw7d7v8gwz.1 and adds the extended validation data and documentation. It is a separate dataset record rather than a new version under the original DOI. The archive contains data and documentation; analysis and solver source code are not included.
Files
Steps to reproduce
1. Extract the ZIP archive and read the included README for the file inventory, variable definitions, and correspondence with the manuscript and supplementary material. 2. Load the .npy files using NumPy with allow_pickle=False and read the JSON files using UTF-8 encoding. Follow the documented indexing conventions and distinguish unspecified trust entries from observed zeros. 3. Use the archived inputs, parameter settings, random seeds, and solver metadata to configure the models described in the manuscript. Apply the documented cost-scaling and attribute-weighted consensus conventions. 4. Inspect the archived run records and summary files to examine clustering outputs, consensus trajectories, adjustment costs, out-of-sample evaluations, and scale experiments. Distinguish formal scale runs from serial pilot runs and retain the recorded solver termination outcomes. 5. Use the SHA-256 manifest to verify file integrity. The archive supports inspection and reuse of the numerical inputs and recorded outputs; rerunning the optimization requires an implementation of the manuscript's models.
Institutions
- Nanjing University of Information Science and TechnologyJiangsu, Nanjing