Semantic Feature Lineage and Prediction–Explanation Compatibility Verification in Cloud-Based Big Data Systems

Anurag Kumar Joshi

Semantic Feature Lineage and Prediction–Explanation Compatibility Verification in Cloud-Based Big Data Systems

Author(s)	Mr. Anurag Kumar Joshi
Country	India
Abstract	Abstract A numerically plausible explanation can describe the wrong record, transformed feature or model output when prediction and explanation artifacts are assembled across distributed pipeline stages. Conventional provenance supports reconstruction, but retaining identifiers alone does not establish that an explanation is compatible with the prediction being reported. This exploratory architecture paper develops a semantic feature-lineage register and a two-stage compatibility gate for tabular classification workflows designed for cloud deployment. The register captures source-variable relationships, fitted transformation parameters, ordered model coordinates, units and reporting rules. The gate verifies request dependencies before generation and binds the returned explanation to its prediction before accepted release. The study combines a formal contract, two exact counterexamples and descriptive reanalysis of the uploaded doctoral thesis; it does not claim a newly executed cloud benchmark. The first counterexample proves that a metadata-only verifier cannot distinguish payloads sharing identical metadata. The second shows that equal prediction values cannot establish correct feature association. A further accounting condition identifies when grouped transformed contributions preserve their total without establishing original-variable Shapley semantics. The thesis reports 4,980 detections among 5,000 injected mismatches, giving a recalculated detection rate of 99.60%, and twelve false rejections among 100,000 valid controls, giving 0.012%. These finite-suite summaries are not independently authenticated, and architecture-specific improvements and overhead remain unresolved. The contribution is a release-time compatibility specification with explicit trust boundaries, failure states and an independently assessable fault-testing design. It separates association integrity from explanation fidelity, predictive quality and causal interpretation. Keywords: semantic feature lineage; explainable artificial intelligence; provenance; compatibility verification; cloud computing; fault injection; prediction–explanation integrity.
Keywords	Abstract A numerically plausible explanation can describe the wrong record, transformed feature or model output when prediction and explanation artifacts are assembled across distributed pipeline stages. Conventional provenance supports reconstruction, but retaining identifiers alone does not establish that an explanation is compatible with the prediction being reported. This exploratory architecture paper develops a semantic feature-lineage register and a two-stage compatibility gate for tabular classification workflows designed for cloud deployment. The register captures source-variable relationships, fitted transformation parameters, ordered model coordinates, units and reporting rules. The gate verifies request dependencies before generation and binds the returned explanation to its prediction before accepted release. The study combines a formal contract, two exact counterexamples and descriptive reanalysis of the uploaded doctoral thesis; it does not claim a newly executed cloud benchmark. The first counterexample proves that a metadata-only verifier cannot distinguish payloads sharing identical metadata. The second shows that equal prediction values cannot establish correct feature association. A further accounting condition identifies when grouped transformed contributions preserve their total without establishing original-variable Shapley semantics. The thesis reports 4,980 detections among 5,000 injected mismatches, giving a recalculated detection rate of 99.60%, and twelve false rejections among 100,000 valid controls, giving 0.012%. These finite-suite summaries are not independently authenticated, and architecture-specific improvements and overhead remain unresolved. The contribution is a release-time compatibility specification with explicit trust boundaries, failure states and an independently assessable fault-testing design. It separates association integrity from explanation fidelity, predictive quality and causal interpretation. Keywords: semantic feature lineage; explainable artificial intelligence; provenance; compatibility verification; cloud computing; fault injection; prediction–explanation integrity.
Field	Engineering
Published In	Volume 8, Issue 6, November-December 2026
Published On	2026-11-10

View / Download PDF File

About IJFMR Fees & Payment Current Issue Publication Archive	Submit Research Paper Track Submission Status Publication Guidelines Publication Ethics Peer Review & Plagiarism	Join as a Reviewer Editors & Reviewers Reviewer Referral Program Get Reviewer Membership Certi.	Website/Journal Policies Usage Policy Content Policies Privacy Policy

Contact Us		+91-9687-828-838	editor@ijfmr.com

International Journal For Multidisciplinary Research

A Widely Indexed Open Access Peer Reviewed Multidisciplinary Bi-monthly Scholarly International Journal

Semantic Feature Lineage and Prediction–Explanation Compatibility Verification in Cloud-Based Big Data Systems

Share this