Audit-Grade Real Estate Transaction Deep Insight Framework

This dataset is intended to support two core research applications: 1. An Audit-Grade Deep Insight Framework for Real Estate Transactions: A Governance-Oriented Evidence-Chain Study of Real Estate Transactions; and 2. An Audit-Grade Deep Insight Framework for Real Estate Transactions: A Quantitative Study Integrating Data Quality Drift Auditing, Quantile Conformal Prediction, Dual-Layer Anomaly Alerting, and Explainable Attribution. Built on the Connecticut Open Data real estate transaction records, the dataset is organized around a central governance question: whether an analytical system still retains governance eligibility under changing data conditions. Rather than limiting real estate analytics to conventional point-accuracy comparison, this dataset is designed to support a broader institutional framework centered on governance adoptability, auditability, and traceable risk escalation. The dataset is structured as a governance-oriented evidence chain covering five interconnected analytical layers: data quality drift auditing, quantile conformal prediction, transaction-level anomaly detection, system-level alerting, and explanation stability auditing. Its purpose is not merely to identify which model performs better, but to determine which forms of evidence remain organizationally adoptable when distributions continue to shift, local anomalies accumulate, and explanatory outputs must enter formal review procedures. In this sense, the dataset is designed to support decisions about when to refresh decision baselines, when to escalate local deviations into systemic warning signals, and when model explanations are sufficiently stable to enter an audit loop. Accordingly, this dataset should be understood not only as a real estate transaction data resource, but also as an audit-grade, governance-oriented deep-insight reproducibility package. It is suitable for research and practice in real estate analytics, risk alert modeling, explainable AI auditing, organizational governance, and open-science reproducibility, especially for demonstrating how drift auditing, conformal interval prediction, dual-layer anomaly alerting, and explainable attribution can be integrated into a unified quantitative framework with governance, audit, and managerial relevance.