Input: URL, domain_authority (or similar), anchor_text, link_type (dofollow/nofollow), ref_domain_count, etc.
Output: Quality score (0–1) and/or risk score (spam likelihood); optional binary keep/disavow.
Models: XGBoost/LightGBM on tabular features; optional text embedding for anchor or URL for spam detection.

Tech Stack

Category	Tools
ML	scikit-learn, XGBoost, LightGBM
Data	pandas, NumPy
Optional NLP	sentence-transformers (anchor/URL)

Setup

pip install -r requirements.txt

Usage

bash

python train.py
python inference.py --input data/backlinks.csv --output scored_links.csv

Project structure

10_backlink-quality-scorer/
├── config.py
├── train.py           # Quality (regression) and/or risk (classification)
├── inference.py       # Add pred_quality_score, pred_risk_label
├── requirements.txt
├── .env.example
├── data/
│   └── backlinks.csv      # Sample: features + quality_score, risk_label
└── models/

Data

Sample data (included): data/backlinks.csv — columns: url, domain_authority, dofollow, ref_domains, anchor_text (optional; anchor_length is derived), same_topic; targets: quality_score (0–1), risk_label (0/1).
Set DATA_PATH in .env if using another file.

License

MIT.

alirezaaminzadeh/backlink-quality-scorer

Backlink-Quality-Scorer: Backlink Quality and Risk Scoring

Author

Problem

Approach

Tech Stack

Setup

Usage

Project structure

Data

License

alirezaaminzadeh/
backlink-quality-scorer