End-to-end Customer Segmentation Project

Dataset

The data is downloaded from Kaggle.

Tools Used in This Project

• Prefect: Orchestrate workflows
• hydra: Manage configuration files
• pre-commit plugins: Automate code reviewing formatting
• poetry: Python dependency management
• DVC: Data version control
• BentoML: Deploy and serve machine learning models

Project Structure

• src: consists of Python scripts
• config: consists of configuration files
• notebook: consists of Jupyter Notebooks
• tests: consists of test files

Set Up the Project

1. Install Poetry
2. Set up the environment:

make setup

3. To persist the output of Prefect's flow, run

export PREFECT__FLOWS__CHECKPOINTING=true

Run the Project

To run all flows, type:

python src/main.py

To run the process flow, type:

python src/main.py flow=process

To run the segment flow, type:

python src/main.py flow=segment

Serve Machine Learning Models with BentoML

To serve the trained model, run:

bentoml serve src/bentoml_app.py:service --reload

To send requests to the newly started service in Python, run:

python src/predict.py

Details of predict.py:

import requests

prediction = requests.post(
    "http://127.0.0.1:5000/predict",
    headers={"content-type": "application/json"},
    data='{"Income": 58138, "Recency": 58,"MntWines": 635,"MntFruits": 88,"MntMeatProducts": 546,"MntFishProducts": 172,"MntSweetProducts": 88,"MntGoldProds": 88,"Complain": 0,"Response": 1,"age": 64,"enrollment_years": 10,"family_size": 1}',
).text

print(prediction)

Output:

2