Acme Security is a fictional company operating across the Gulf and Middle East region. This project demonstrates a safe, generic data science workflow to predict whether a customer will convert based on recent engagement signals. All data and logic are synthetic and safe for public use.
This repo focuses only on data science modeling and analysis. It does not include ETL, warehousing, or production pipeline code.
Synthetic CSV with aggregated customer engagement metrics:
- Geography (country, city)
- Channel and device type
- Recent activity (visits, events, recency)
- Target label
converted(0/1)
See docs/data_dictionary.md for field definitions.
- Load synthetic dataset
- Basic EDA (optional notebook)
- Train a baseline classification model
- Evaluate with accuracy and F1 score
- Python 3
- pandas
- scikit-learn
- Create and activate a virtual environment.
- Install dependencies:
pip install -r requirements.txt - Train the model:
python src/train.py
make install, make train
The training script prints metrics like:
Model metrics:
accuracy: <value>
f1: <value>
Values depend on the synthetic data and split.
Everything in this repository is fictional and synthetic. No real company names, schemas, data, or business logic are included.
- Add cross‑validation and model persistence
- Add simple feature importance report
- Add automated data checks
MIT