Website Technology Evidence Classifier
A scikit-learn text classifier trained on the companion Website Technology Evidence Dataset.
Method
A FeatureUnion combines word-level TF-IDF (unigrams/bigrams) and character-level TF-IDF (3–5 grams), followed by Logistic Regression.
The character features improve tolerance for URLs, paths, version strings and small formatting differences.
Evaluation
See metrics.json.
Important: the companion dataset is synthetic and fingerprint-driven. Evaluation scores measure performance on this controlled benchmark and must not be presented as accuracy on arbitrary real-world websites.
Output
inference.py returns the top five predicted technologies with confidence scores.
Related project
Website Technology Detector — AzkiWeb
https://azkiweb.com/
https://azkiweb.com/website-design-price
MIT License.