Module confidence 

Confidence scoring module — transforms raw dependency data into actionable intelligence.

Implements three approaches from the research synthesis:

1. Noisy-OR fusion (Google Knowledge Vault, KDD 2014) — for multi-source evidence
2. Temporal decay (TempValid framework, ACL 2024) — confidence is not timeless
3. Bayesian update — for incorporating new evidence

A static dependency graph says “A depends on B.” A probabilistic one says “A depends on B with 0.87 confidence based on static analysis (0.92) and test coverage (0.75), decaying at 0.01/day since last verification.”

Structs

ConfidenceReport

Confidence report for explaining to humans.

SourceContribution

Functions

bayesian_update

Simple Bayesian update: combine prior belief with new evidence.

chain_confidence

Propagate confidence through a transitive chain.

multi_path_confidence

Combined multi-path confidence: given multiple paths between A and C, compute the overall confidence using noisy-OR over each path’s chain confidence.

noisy_or

Noisy-OR model for combining independent evidence sources.

temporal_decay

Temporal confidence decay using exponential model.