ML Project Workspace: Churn, Loan, Banking & Flight Prediction

Gabriel Adebayo

ML Engineer

Keras

TensorFlow

XGBoost

Artificial Intelligence

📊 ML Project Workspace – Churn, Loan, Banking Market & Flight Prediction

This repository contains structured work on real-world ML problems including Customer Churn, Loan Prediction, Banking Market Response, and Flight Price Prediction. It focuses on robust data pipelines, exploratory data analysis (EDA), preprocessing, model evaluation using multiple metrics with recall emphasized for imbalanced scenarios.

🛠 This workspace is designed to evolve. New datasets will follow a reusable pipeline and evaluation logic for consistency and clarity.

✅ Current Datasets

1. 📉 Customer Churn Prediction

Goal: Identify customers likely to churn.

Key Steps:

Cleaned and mapped binary features

OneHotEncoded categorical variables

Scaled features with StandardScaler

Balanced data using SMOTE

Trained:

Logistic Regression

Random Forest

AdaBoost

Gradient Boosting

XGBoost

Evaluated with:

Accuracy

Precision, Recall

F1 Score

ROC AUC

Focus: Maximize recall to catch at-risk customers

2. 💰 Loan Prediction

Goal: Predict loan approval status.

Key Steps:

EDA on categorical and numeric features

Handled missing values, label/one-hot encoding

Compared classifiers with emphasis on recall

Best Model: Logistic Regression (good balance of interpretability and performance)

3. 🏦 Banking Market Response (Term Deposit Subscription)

Goal: Predict if a customer will subscribe to a term deposit.

Key Steps:

Cleaned and encoded socio-economic features

Explored job, education, and campaign-related behaviors

Balanced the dataset using SMOTE

Focused on avoiding false negatives (missed interested clients)

Compared models using recall, F1-score, ROC AUC

4. ✈️ Flight Price Prediction (Deep Learning)

Goal: Predict flight ticket prices based on travel information using Deep Learning.

Key Steps:

Extracted features from date/time columns

Converted text data (e.g. airlines, source, destination) using encoding

Scaled continuous variables

Trained regression models including:

Linear Regression

Random Forest Regressor

XGBoost Regressor

✅ Neural Network (Deep Learning) using Keras/TensorFlow

Metrics Used:

MAE, MSE, RMSE, R² Score

Emphasis: Low RMSE and strong generalization using DL model

🔍 Evaluation Highlights

Problem Priority Metric Why It Matters Churn Recall Catch all at-risk customers Loan Recall Avoid missing eligible applicants Bank Recall Don’t miss likely subscribers Flight RMSE / R² Accurate price estimation using deep learning

🧪 How to Run

# Activate your environment
conda activate mlenv  # or virtualenv

# Run Jupyter
jupyter notebook

🧭 Project Structure

📁 project_root/
│
├── 📄 README.md <- This file
│
├── 📁 TeleCustomerChurnNotebook/ <- VScode folder with scripts, models, and EDA
│
├── 📁 LoanPredictionProblem/ <- VScode folder with scripts, models, and EDA
│
├── 📄 BankMarketing.ipynb <- Google Colab notebook (single file)
├── 📄 Flight_Prediction.ipynb <- Google Colab notebook (DL-based)
│
└── 📄 requirements.txt <- Project dependencies

🚀 What's Next

Add datasets in health, fraud detection, and education

Modularize reusable preprocessing and training pipelines

Convert selected notebooks into end-to-end ML/DL apps

Version experiments and models using MLflow

Add Docker & cloud deployment support (AWS, Azure, GCP)

🙌 Acknowledgments

This project is part of my ML Engineering journey — building toward deployable, production-ready solutions using best practices in data science and MLOps.

🔄 Feedback, suggestions, or collaborations are always welcome!

Like this project

Posted Sep 1, 2025

Developed ML models for churn, loan, banking, and flight prediction.

Likes

Views