Bank Marketing Dataset - MLflow Project

1-Table of Contents
2-Introduction
3- Dataset Description
4- Features
5- Setup and Installation
6- Project Workflow
7- MLflow Integration
8- Results and Evaluation
9- Contributing
10- License

Introduction

The Bank Marketing Dataset MLflow Project is a machine learning project that predicts whether a client will subscribe to a term deposit (deposit as target variable) based on their demographic and interaction data. This repository utilizes MLflow to streamline experiment tracking, reproducibility, and model deployment.

Dataset Description

The dataset used in this project is from the Bank Marketing Dataset on Kaggle.

• Source: UCI Machine Learning Repository
• Size: ~45,000 rows and 17 features
• Objective: Predict the outcome of the marketing campaign (deposit: yes/no).

Features

Key features include:

• Demographic Information: age, job, marital, education.
• Campaign Details: campaign, pdays, previous, poutcome.
• Financial Data: balance, loan, housing.
• Date Information: month, day_of_week.

Setup and Installation

Prerequisites

• Python 3.8 or later
• MLflow installed
• Libraries: pandas, scikit-learn, xgboost, seaborn, matplotlib

Project Workflow

1. Exploratory Data Analysis (EDA):

• Visualize distributions, correlations, and outliers.
• Tools: seaborn, matplotlib.

2. Preprocessing:

• Handle missing data, encode categorical features, and scale numerical ones.
• Techniques: LabelEncoding, OneHotEncoding, StandardScaler.

3. Model Training:

• Models used: Logistic Regression, Random Forest, XGBoost.
• Feature selection and hyperparameter tuning.

4. Evaluation:

• Metrics: Accuracy, Precision, Recall, ROC-AUC.

5. Experiment Tracking:

• Log parameters, metrics, and artifacts in MLflow.

6. Model Deployment:

• Save the best-performing model for deployment.

MLflow Integration

What is MLflow? MLflow is an open-source platform for managing the end-to-end machine learning lifecycle:

• Tracking: Log metrics, parameters, and models.
• Projects: Reproducible packaging of code.
• Models: Deployment and sharing of models.
• Registry: Centralized model store.

Links:

• Project Notebook
• Dataset