MlopsZoomcampCourceProject 2022

Repository includes code and document related to project for MLOps Zoomcamp course https://github.com/DataTalksClub/mlops-zoomcamp

Problem statement: Credit Card Attrition

Overview

• Attrition on credit cards has become a major worry for the banking industry. Given that the expense of obtaining new consumers is higher than the expense of keeping existing customers, it significantly affects profitability.
• In this work, a selection of supervised Machine Learning models to identify which customers want to cancel their credit cards is evaluated.
• Attrition prediction models are important to reduce the cancellation of credit card products and analyze which customers have a greater tendency to cancel.
• This work was carried out through a strategic segmentation focusing on the client to generate campaigns according to the profile of each one, to understand and anticipate their behavior. Likewise, productive resources were focused on high-value groups, as it is cheaper to retain a client than attract a new one.

Objective

Predicting customers who are likely to drop off credit card. to work on customer retaintion strategy like running compaigns.

Dataset

Credit Card Churn Prediction used for this project.

Directory info: each directory have own readme file that talk about instruction to run

• /web-service: build flask api + docker pakaging + deployment api through docker

• /lambda-terraform: build lambda function + docker pakaging + terraform LaC + Mlflow

• /monitoring : build batch monitoring service with evidently

note : please check my "feature" branch for below change https://github.com/vermadev54/MlopsZoomcampCourceProject/tree/feature

• /allconcept : build streaming api with (unit test + code lint + LaC + cicd )

Solution i have build random forest classification model with accuracy ~98%

• Model training

python3 train.py --data_path 

• data preprocessing/training/mlflow experiment tracking /prefect Workflow orchestration

python3 preprocessing_data.py

• Model registery

python3 registry_model.py --data_path ./output

• Prefect deployment with scheduled crons

prefect deployment create prefect_deploy.py

running mlflow server with cloud db and s3 storage

mlflow server -h 0.0.0.0 -p 5000 --backend-store-uri postgresql://mlflow:FdXoiuOCyQvyiDL0Gftk@mlflow-database.ciuzmsnp32jg.us-east-1.rds.amazonaws.com:5432/mlflow_db --default-artifact-root s3://jai-mlops-zoomcamp-tfstate

running prefect server on remeote server for Workflow orchestration

prefect config unset PREFECT_ORION_UI_API_URL
prefect config set PREFECT_ORION_UI_API_URL="http://54.146.195.209:4200/api"
prefect orion start --host 0.0.0.0