Work Description

Date: 21 March, 2025

Dataset Title: Data and code for paper title "Bayesian Model Selection for Network Discrimination and Risk-informed Decision Making in Material Flow Analysis"

Dataset Contact: jkliao@umich.edu

Dataset Creators:
Name: Jiankan Liao
Email: jkliao@umich.edu
Institution: University of Michigan Department of Mechanical Engineering
ORCID: https://orcid.org/0000-0001-8104-8388

Name: Sidi Deng
Email: sidideng@umich.edu
Institution: University of Michigan Department of Mechanical Engineering
ORCID: https://orcid.org/0009-0003-9234-1952

Name: Xun Huan
Email: xhuan@umich.edu
Institution: University of Michigan Department of Mechanical Engineering
ORCID: https://orcid.org/0000-0001-6544-2764

Name: Daniel R Cooper
Email: drcooper@umich.edu
Institution: University of Michigan Department of Mechanical Engineering
ORCID: https://orcid.org/0000-0003-2903-0468

Funding:
- National Science Foundation (NSF), under the Award Number 2040013

Nomenclature:

Key Points:
- The code supporting the paper "Bayesian Model Selection for Network Discrimination and Risk-informed Decision Making in Material Flow Analysis"

Research Overview:
We apply Bayesian inference to reduce network structure uncertainty on material flow analysis (MFA) and demonstrate the methodology through a case study on U.S. steel flow. In addition, we derive an input/output-based analysis to conduct decision-making based on the uncertainty results from MFA

Methodology:
We apply sequential Monte Carlo through the python pymc3 package to generate posterior samples and marginal likelihood used to quantify MFA parametric and network structure uncertainty. We compile our own rectified input/output methods to derive the envrinoemntal impacts associated with the consumption nodes in the steel industry case study.

Instrument and/or Software specifications: pymc3 package for conducting Bayesian inference

Files contained here:
(1)--(16): Inference for **** model.ipynb
-These 16 files are the code used to generate posterior samples as well as the marginal likelihood corresponding to specific MFA network structure for the case study

(17): load pickles-bayesian averaging model.ipynb
-This file contains the code to derive the Bayesian averaged prior and posterior predictive mass flows for the case study as well as to conduct decision making using the rectified input/output methods

(18): Data_for_figure.xlsx
-This file contains the underlying data to create the figures in the manuscript and the supporting information of the paper

Related publication(s): Liao, Jiankan, Deng, Sidi, Xun Huan, and Daniel Cooper. "Bayesian Model Selection for Network Discrimination and Risk-informed Decision Making in Material Flow Analysis." arXiv preprint arXiv:2501.05556 (2025).

Use and Access:
This data set is made available under an Attribution 4.0 International (CC BY 4.0).

To Cite Data:
Liao, Jiankan, Deng, Sidi, Xun Huan, and Daniel Cooper. "Bayesian Model Selection for Network Discrimination and Risk-informed Decision Making in Material Flow Analysis." arXiv preprint arXiv:2501.05556 (2025).