Method

Dorn

We reconstruct the Dorn model from the public feature data and released visual-metric implementations. Metrics available for fewer than 80% of training samples are excluded; forward selection retains seven features, and logistic regression is fitted on all 360 Dorn samples. A thin adapter selects the appropriate language analyzer without changing metric formulas or wrapping Java snippets. The score is the readable-class probability.

Score scale: 0–1; higher is more readable.

Reproduction code and instructions

Dataset results