Universal Bank Loan Prediction: k-NN Model

This repository contains the machine learning project that employs k-NN classification to predict if a Universal Bank customer will accept a personal loan offer.

Table of Contents

1. Submission Instructions
2. Data Description
3. Scenario
4. Tasks

Submission Instructions

• Answers should be written as text or comments within the R codes.
• Compile R scripts to html/word/pdf for submission.
  • Shortcut: Ctrl + Shift + K then select "html".
• Submit the html file via Canvas before the deadline.

Data Description

• Dataset: Universalbank.csv.
• Contains data on 5000 bank customers.
• Includes demographics, bank relationship details, and past loan campaign responses.
• Key metric: Only 9.6% accepted the last personal loan offer.

Scenario

Universal Bank seeks to convert its liability customers to personal loan customers. With a previous campaign seeing a 9% conversion rate, the bank aims to optimize its strategy using k-NN predictions for better-targeted campaigns.

"Do the right thing; and do it right." - Master Yeoda.

Tasks

1. Data Handling
   • Import and split data: 60% training and 40% validation. a. Check variable data types. b. Set up kNN with relevant variables.
2. k-NN Modelling
   • Exclude ID and ZIP code predictors.
   • Factorize categorical variables.
   • Test with k values: 3, 5, and 7. Select the best k.
3. Model Assessment
   • Predict on the validation set using optimal k. a. Visualize with an ROC curve. b. Gauge model efficacy.
4. Sample Prediction
   • Predict loan acceptance for a specific customer profile provided.