Predicting Term Deposit Subscriptions at a Bank

A supervised machine learning project predicting whether a customer will subscribe to a term deposit based on demographic and campaign data.

---

Overview

Developed an ML model to assist marketing teams in identifying customers most likely to respond positively to campaigns.

---

Steps

1. Data Exploration

   • Performed EDA on both categorical and numerical data to gain insights into the variables.
   • Utilized the insights to clean/preprocess the data.
   • Visualised the response variable to understand the class imbalance.

2. Data Preprocessing

   • Cleaned and encoded customer's demographic attributes (age, job, education, marital status, etc.).
   • Handled outliers and cleaned numerical data.
   • Handled missing values and categorical variables using pandas and numpy.

3. Dealing with Class Imbalance

   • Applied SMOTE (Synthetic Minority Over-sampling Technique) to balance data the negative responses to positive responses.

4. Modeling

   • Implemented a Logistic Regression, Random Forest Classifier, Ada Boost Classifier.
   • Compared the evaluation metrics to each other and selected the Random Forest Classifier (best performance).
   • Tuned hyperparameters using GridSearchCV.

5. Evaluation

   • Achieved:
     • AUC: 0.93
     • Precision: 0.90
     • Recall: 0.89