Bank Marketing Campaign Optimization using Machine Learning
Project Overview
This project builds an industry-level machine learning pipeline to predict whether a customer will subscribe to a term deposit. The goal is to optimize marketing campaigns by identifying high-probability customers while minimizing unnecessary calls.
Dataset
Source: UCI Machine Learning Repository
Dataset: Bank Marketing (bank-additional-full.csv)
Records: 41,188 customers
Business Objective
- Increase subscription conversion rate
- Reduce marketing cost
- Improve targeting efficiency
Models Used
- Decision Tree (Baseline)
- Random Forest (Final Model)
Final Model Performance (Random Forest)
- Accuracy: ~87%
- Recall (Subscribed): 68%
- ROC-AUC: 0.76
- Threshold Optimization: 0.4 (Business Balanced Strategy)
Key Insights
- Economic indicators strongly influence subscription behavior.
- Previous successful campaigns increase probability of conversion.
- Excessive campaign contact reduces effectiveness.
- Age and employment conditions affect customer decisions.
Business Impact
Using threshold tuning:
- Improved customer detection from 25% → 68%
- Increased projected campaign profitability
- Reduced unnecessary marketing calls
Technologies Used
- Python
- Pandas
- Scikit-Learn
- Matplotlib
- Seaborn
Nayana Goudar