📊 Term Deposit Subscription Prediction – Bank Marketing Dataset

This project uses classification models to predict whether a customer will subscribe to a term deposit based on a variety of demographic and campaign-related factors.

🧠 Objective

To build a predictive model using the Bank Marketing dataset (2015–2017) to support marketing decision-making through data-driven insights.

🛠️ Tools Used

• Python
• Pandas, NumPy
• Scikit-learn
• Matplotlib, Seaborn

🔍 Key Techniques

• Exploratory Data Analysis (EDA)
• Feature engineering and scaling
• Handling class imbalance with class_weight='balanced'
• Logistic Regression, Decision Tree, and Random Forest models
• Model evaluation using classification report, confusion matrix, and ROC-AUC

📊 Visuals

• Subscription class distribution

• Confusion Matrices

Logistic Regression

• Decision Tree

• Random Forest

  

📈 Performance Snapshot - Classification report

📁 Files

• Term_Deposit_Prediction_BankMarketing.ipynb – Full Jupyter notebook with code and results
• bank.csv – Source dataset (not uploaded here for privacy)

🚀 How to Run

pip install pandas numpy scikit-learn matplotlib seaborn

Then open the notebook in Jupyter or Google Colab and run all cells.

🧠 Interpretation of Results

🎯 Business Context

The goal was to predict whether a customer would subscribe to a term deposit offer, based on demographic and campaign-related features. The challenge: only a small percentage of customers said “yes” — making it a classic imbalanced classification problem.

📈 Model Findings

• Logistic Regression achieved the best balance for this use case with 62% recall, ideal for identifying likely subscribers.

• Decision Tree performed well in terms of accuracy but had lower recall, limiting its usefulness for identifying "yes" customers.

• Random Forest had the highest precision and accuracy but the lowest recall (11%), meaning it missed most of the actual subscribers — not ideal for this use case.

🔍 Key Factors Driving Subscription

• Contact method (cellular)
• Education level
• Credit default history

Note: duration was excluded to prevent data leakage.

✅ What This Means

• Logistic Regression or an ensemble method that prioritizes recall is recommended when the goal is to identify as many potential subscribers as possible..

• These models help the bank:

  • Prioritize high-potential customers.
  • Tailor campaigns to effective segments.
  • Reduce wasted effort and cost

📌 Insights & Conclusion

• The project demonstrates how different models can trade off recall vs. precision — an important business decision point.
• Logistic Regression, while simple, achieved the best trade-off between recall and precision.
• Key features like contact method, education, and loan history were the strongest predictors of term deposit subscription.
• These models can help the bank prioritize outreach, refine targeting, and increase campaign efficiency.