Data for: Semantic-Enhanced Assisted Evacuation Decision-Making Simulation in Nursing Homes Using Knowledge Graphs

Published: 16 November 2026| Version 5 | DOI: 10.17632/4ckxvmsfhd.5
Contributors:
,
,
,
,

Description

This dataset supports the manuscript “Semantic-Enhanced Assisted Evacuation Decision-Making Simulation in Nursing Homes Using Knowledge Graphs”. It contains the data, source code, and supporting materials used to construct, execute, and verify the revised knowledge graph-driven assisted-evacuation framework. The dataset includes the formal evacuation ontology, portable Neo4j knowledge-graph artifacts, Resolver and simulation source code, scenario and parameter definitions, and the formal results for BASE, STAFF1, STAFF2, MOB1, READY1, and the five single-factor sensitivity conditions (V1–V5) examined in Section 5.3. Across these ten conditions, it contains 300 materialized ResolvedScenario files and results from 1,200 formal simulation runs, with 30 paired random seeds and four dispatch strategies per condition. The original five conditions are documented by individual formal simulation records, while the 600 additional sensitivity runs are provided in a complete run-level results table. Statistical-analysis code and a portable reproduction adapter for the original five conditions are provided together with frozen reference outputs and verification records. The sensitivity addendum includes the actual scenario inputs, condition definitions, statistical-analysis protocol, summary and full statistical results, and verification records supporting Section 5.3 and Table 8. The existing STAFF2 results serve as the sensitivity reference and are reused without additional baseline runs. Source data and native OriginPro project files for the reported figures are also included. README, manifest, checksum, dependency, and verification files document the software environment, package structure, execution procedures, and reproducibility scope.

Files

Steps to reproduce

1. Extract NursingHome_KG_Evacuation_Data(2026.11.15).zip and open the extracted RESS_revision_reproducibility directory. All paths below are relative to this directory. Read the main README.md and section_5_3_sensitivity/README.md for the package structure, software environment, execution procedures, and reproducibility scope. 2. Inspect kg/ontology/evac_core_v2.ttl and the files in kg/base to examine the formal ontology, instantiated knowledge graph, Neo4j import/verification scripts, and competency queries. 3. Use the materialized scenarios in resolved and the source code in simulation and experiments to inspect or execute individual formal simulations for BASE, STAFF1, STAFF2, MOB1, and READY1, following the environment setup in the main README. Each condition contains 30 paired seeds (0–29) evaluated under strategies S1–S4. 4. For the five sensitivity conditions in Section 5.3, consult section_5_3_sensitivity/conditions.csv for the mapping between V1–V5 and their file identifiers. The scenarios subdirectory contains 150 additional materialized inputs, covering five conditions and 30 seeds; each input is shared by S1–S4. These inputs can be supplied to the same public simulation runner. The existing STAFF2 scenarios and results provide the reference condition. 5. To reproduce the statistics for the original five conditions, use Python 3.12.14 and the dependencies specified in requirements.txt. From the root of a fresh extracted working copy, run: python -B reproduction/reproduce_statistics.py The reproduced statistics are written only to reproduction/output. Confirm successful reproduction using the generated verification receipts. This adapter was validated against the stored statistical results for BASE, STAFF1, STAFF2, MOB1, and READY1 with exact numerical agreement; it does not process V1–V5. 6. To examine the sensitivity results supporting Section 5.3 and Table 8, use section_5_3_sensitivity/data/metrics_600.csv for the 600 run-level observations. Consult analysis/analysis_protocol.json for the statistical methods, analysis/compact_results.csv for summary results, and analysis/statistical_results.json for the complete saved statistical outputs. The accompanying README explains how these outputs map to Table 8, and verification/statistical_validation.json records the historical independent validation. 7. To regenerate Fig. 4 and Fig. 5, open the supplied native .opju projects in OriginPro 2024 and use the accompanying source data and figure-regeneration notes.

Categories

Computer Science, Engineering, Fire Safety, Building Simulation, Crowds Simulation

Funders

  • National Natural Science Foundation of China
    Beijing, Beijing
    Grant ID: 72671030; 72274123; 72174099; 72674189
  • Beijing Municipal Natural Science Foundation
    Grant ID: 8242014
  • National Key R&D Program of China
    Grant ID: 2024YFC3017000
  • Civil Aircraft Fire Science and Safety Engineering Key Laboratory of Sichuan Province
    Grant ID: MZ2025KF01

Licence