PASY IMPLEMENTATION DATA

This dataset is the supplementary data for the manuscript “Performance Modeling for Isolated Industrial Operations: A Data Envelopment Analysis Framework” (Inana, O.), submitted to Operations Research Forum. It contains the input data, configuration, intermediate outputs and results of the Pernage Analytics System (PASy). PASy is a framework that applies bounded Data Envelopment Analysis (DEA) to a single facility with no external peers, by building synthetic comparators from the facility’s own daily operating data. Data. The daily operating records are fictional. They were generated with a large language model under constraints designed by the author, and they describe no real facility. The results show how the framework behaves, not how any real asset performs. Contents. One Excel workbook, THE_EXTENDED_PASY_FILE_3.xlsx, with 15 sheets: README: guide to the sheets AGGREGATED_DAILY_REPORTS: 365 daily records of 20 operational variables (input data) SURVEY_INPUT, AHP_PAIRWISE_COMPARISON_MATRIX, AHP_WEIGHTS: the expert survey, the agreed pairwise judgements, and the resulting weights and consistency ratio PRISM_CLASSIFICATION_FRAMEWORK: variable roles, expert ratings, multipliers, Beta parameters and band widths SYNTHETIC_DMUS: the 21,900 synthetic peers of the baseline run COMPUTATION_SHEET: daily optimistic, pessimistic and geometric efficiency scores PEER_PROFILE_LOG: scores of the 20,160 same-day synthetic peers EFFICIENCY_RATINGS: the quartile class of each of the 336 evaluated days SENSITIVITY_RUNS and SEED_RUNS: the sensitivity analyses, specification checks and five-seed replication VARIABLE_RANGES: observed and synthetic range of each variable RESULTS: every statistic reported in Section 4 of the manuscript, as key–value pairs RUN_CONFIG: seed, parameters, physical caps, DEA input and output roles, and hashes identifying the input data and code version Code. The source code of the implementation is not released. The method is described in the manuscript, and RUN_CONFIG records the code version that produced these results. The scripts that draw the manuscript’s figures are available separately on GitHub. Version 3. This version replaces Version 2. All results were recomputed with the revised implementation described in the manuscript’s second revision, and the sheets were reorganised and documented. Cite this version for the revised manuscript.