Chrun case study

Predict customer churn for TELCO Inc and recommend personalized discounts to maximize future profit using a provided dataset.

Project Organization

├── LICENSE            <- Open-source license if one is chosen
├── Makefile           <- Makefile with convenience commands like `make data` or `make train`
├── README.md          <- The top-level README for developers using this project.
├── data
│   ├── external       <- Data from third party sources.
│   ├── interim        <- Intermediate data that has been transformed.
│   ├── processed      <- The final, canonical data sets for modeling.
│   └── raw            <- The original, immutable data dump.
    └── results        <- result csv on the test set.
│    
├── docs               <- A default mkdocs project; see mkdocs.org for details
│
├── models             <- Trained and serialized models, model predictions, or model summaries
│
├── notebooks          <- Jupyter notebooks. Naming convention is a number (for ordering),
│                         the creator's initials, and a short `-` delimited description, e.g.
│                         `1.0-jqp-initial-data-exploration`.
│
├── pyproject.toml     <- Project configuration file with package metadata for rexel_chrun_case_study
│                         and configuration for tools like black
│
├── references         <- Data dictionaries, manuals, and all other explanatory materials.
│
├── reports            <- Generated analysis as HTML, PDF, LaTeX, etc.
│   └── figures        <- Generated graphics and figures to be used in reporting
│
├── requirements.txt   <- The requirements file for reproducing the analysis environment, e.g.
│                         generated with `pip freeze > requirements.txt`
│
├── setup.cfg          <- Configuration file for flake8
│
└── rexel_chrun_case_study                <- Source code for use in this project.
    │
    ├── common.py      <- Script provides helper functions for trian and eval
    │   
    ├── train.py       <- Script to train models and then use trained models and save them
    │
    └── evaluate.py       <- Script to eval and save results 

---

TELCO Churn Prediction Project

Task Description

This project aims to predict customer churn for TELCO Inc., a telecommunications company. The primary objectives are:

• Rank customers by their probability of churning.
• Determine which customers should be contacted to minimize churn.
• Recommend personalized discounts to maximize future profits.

The results are provided in a CSV file containing churn predictions, recommended customer contacts, and suggested discounts.

Project Structure

• mlflow/: Directory for MLflow experiment tracking and model versioning.
• init.py: Initialization file to make the directory importable as a module.
• common.py: Utility functions for model loading, plotting, and discount recommendation.
• config.ini: Configuration file for project parameters and MLflow settings.
• mlflow_exp.py: Script for managing MLflow experiments, tracking metrics, and saving models.
• predict.py: Script for evaluating the model, generating predictions, and saving results.
• train.py: Script for training the churn prediction model and saving it for future use.

Prerequisites

• Python 3.8 or later
• pip
• MLflow

Installation

1. Clone the Project Repository:

   git clone https://github.com/AMMISAIDFaical/ML-engineer_chrun.git
   

Here's your content formatted in Markdown:

## Install Dependencies

```sh
pip install -r requirements.txt

Set Up MLflow

1. Ensure MLflow is installed and set up correctly.
2. Configure the config.ini file with the appropriate settings for MLflow tracking.

Setup and Usage

Configuration

• Edit the config.ini file to set up paths and parameters for MLflow and data files.

Data Preparation

• Place the training and test data CSV files in the appropriate directory.

Train the Churn Prediction Model

python train.py

This script will:

• Load and preprocess the data.
• Train the model using the training dataset.
• Save the trained model for future use.

Predict Churn and Recommend Actions

python predict.py

This script will:

• Load the test data.
• Load the trained model.
• Generate predictions on the test data.
• Evaluate the model using ROC AUC score.
• Save the results in a CSV file.

Output CSV

The output file cv_grid_xgb_results.csv will be saved in ../data/results/ and will contain:

• CUSTOMER_ID: Unique identifier for each customer.
• CHURN_Ground_truth_Label: Actual churn status (LEAVE or STAY).
• CHURN_PREDS: Predicted churn status (LEAVE or STAY).
• CHURN_PROBABILITY: Probability of churn.
• CLIENT_TO_CONTACT: Whether the customer should be contacted (YES or NO).
• DISCOUNT: Recommended discount to offer.
• Rank: Rank based on the probability of churn.

Notes

• Ensure that dataset files are correctly formatted and placed in the specified directories.
• The project includes scripts to train, evaluate, and make predictions using the churn prediction model.
• Modify config.ini and scripts as needed to fit your specific setup and requirements.

Requirements

List of required Python packages:

pandas
scikit-learn
numpy
xgboost
mlflow
matplotlib

Metrics

The model's performance is evaluated using the ROC AUC metric.