Bayesian Data-Mining Pipeline for Climate Policy Stringency Decision Support in China

Decision-support systems must convert heterogeneous, sparse, and noisy policy records into auditable monitoring outputs under uncertainty. We develop a Bayesian data-mining pipeline for climate policy stringency decision support in China. The pipeline first audits six annual input signals: carbon-market prices, market breadth, policy-text intensity, administrative force, green finance, and clean-energy policy actions. It then estimates a latent Climate Policy Stringency Index (CPSI) for 2000-2024 with a Bayesian dynamic factor model and stores MCMC posterior draws as reusable computational artifacts. A posterior diagnostic engine exports machine-readable fields for level uncertainty, annual change probabilities, large-increase alerts, common-component shares, acceleration evidence, jump-model comparison, and external process alignment. In the retained artifact, CPSI increases from -2.10 in 2000 to 6.30 in 2024, with more frequent large-increase alerts after 2020. The temporal-pattern diagnostics support a post-2019 acceleration label, but not a statistically necessary one-year jump at annual frequency. The contribution is a workflow and output-schema refinement: heterogeneous policy records are converted into versioned posterior artifacts, rule-based monitoring fields, and auditable diagnostic labels, while causal effect claims remain outside the measurement boundary.