Self-Healing Cloud Infrastructure Systematic Review: Reproducibility Dataset

Abstract

Data and code for the systematic review "Self-Healing Cloud Infrastructure: A Systematic Review of Machine Learning Approaches for Autonomous Detection, Diagnosis, Decision, and Recovery". The review follows PRISMA 2020 and includes 388 studies published from 2022 to 2026, 359 from the main search through seven discovery channels and 29 from a Web of Science search, out of 566 records assessed at full text. The archive holds the coded study table (canonical_studies.csv) with a ledger of every change since the initial coding, the dated decision log, one extraction record per study with the CASP and risk-of-bias (adapted from PROBAST) rationales, the screening decisions for both searches, the second rater's codes and agreement, the checks of the keyword-based counts, the search harvests with bibliographic fields only, the protocol with its dated deviations log, the rubrics and the screening guide, and the analysis and figure scripts. code/run_all.py regenerates every derived table, number and figure from the data. README.md describes each file. This version accompanies the revised article. Earlier versions describe an earlier state of the corpus. Version v3.1 differs from v3 only in protocol/protocol.md, where the protocol's dated sentence on generative AI is kept with a note pointing to the article's declaration.