Project 04IntermediateFinance6-8 focused hoursCredit-risk modeling case study
LearnPracticeBuild

Loan Default Risk Analysis

Explore borrower characteristics and design a structured workflow for predicting loan default risk.

Data CleaningEDAFeature EngineeringClassificationModel Evaluation
FRAMEDATAANALYZEMODELEVALUATECOMMUNICATE

BUSINESS CONTEXT

Why this project matters

A lending team wants a more structured way to identify risky applications before approval. They need both exploratory insight into borrower behavior and a defensible modeling workflow that reflects the cost of bad lending decisions.

PROJECT OBJECTIVE

What you are expected to accomplish

Analyze borrower and loan features, identify the strongest default-risk patterns, and outline or build a classification workflow with metrics suited to lending decisions.

DATASET OVERVIEW

Expected dataset structure

Expect borrower and loan-level data covering demographics, income, credit history, repayment behavior, loan terms, and whether the borrower defaulted.

Main entities / rows

One row per loan application or funded loan

Target variable

defaulted (Yes/No)

loan_id

string

Unique identifier for the loan record.

annual_income

float

Reported borrower income, potentially missing or skewed.

employment_length

categorical

Borrower employment tenure bucket or years.

credit_score_band

categorical

Grouped creditworthiness measure.

debt_to_income_ratio

float

Debt burden relative to income.

loan_amount

float

Requested or approved principal amount.

loan_purpose

categorical

Reason for the loan, such as education or debt consolidation.

defaulted

binary

Target indicating whether the borrower defaulted.

Data quality issues to expect

  • Income and employment history may be missing or inconsistently formatted.
  • Default events are often less common than non-defaults, creating class imbalance.
  • Some variables may only be available after approval or after repayment begins, which creates leakage risk.

QUESTIONS TO ANSWER

Focus the work around meaningful decisions

  1. 01

    Which borrower characteristics are most strongly associated with default?

  2. 02

    How do income, credit score, debt burden, and loan purpose relate to risk?

  3. 03

    Is the default rate imbalanced enough to make accuracy misleading?

  4. 04

    Which engineered features could make risk patterns clearer or more predictive?

  5. 05

    What evaluation strategy would be appropriate for a lending use case?

PROJECT ROADMAP

Move through the case study in a practical sequence

  1. 01

    Step 01

    Understand the lending decision

    Clarify what default means and how predictions would be used in practice.

  2. 02

    Step 02

    Audit the loan dataset

    Inspect class balance, data types, missingness, and candidate leakage fields.

  3. 03

    Step 03

    Clean borrower records

    Handle inconsistent employment, income gaps, and noisy categorical labels.

  4. 04

    Step 04

    Perform risk-focused EDA

    Compare default rates across borrower, loan, and debt profiles.

  5. 05

    Step 05

    Engineer risk features

    Create bins, ratios, or grouped features that reflect credit behavior more clearly.

  6. 06

    Step 06

    Model and evaluate carefully

    Choose metrics and thresholds that reflect the cost of risky false negatives.

  7. 07

    Step 07

    Summarize a credit-risk workflow

    Document findings, caveats, and a practical next step for the lending team.

TASKS / MILESTONES

Concrete work to complete

  • Measure the default rate and assess class imbalance.
  • Inspect missing values in income, employment, and credit-related features.
  • Compare default by credit score band, loan purpose, and debt-to-income range.
  • Create at least two useful engineered features or grouped variables.
  • Train a baseline classifier or outline one clearly if modeling is out of scope.
  • Evaluate results with metrics beyond accuracy and explain why they matter.
  • Document any leakage risks or unrealistic features you exclude.

SUGGESTED VISUALIZATIONS

Visuals worth creating

  • Default-rate bar chart by loan purpose
  • Credit-score band versus default comparison
  • Debt-to-income distribution by default status
  • Confusion matrix for the chosen model threshold
  • Precision-recall curve or metric comparison table

DELIVERABLES

What the learner should produce

  • Cleaned borrower-loan dataset
  • Exploratory risk analysis notebook
  • Feature-engineering notes
  • Baseline model or evaluation workflow
  • Risk findings and lending recommendations

SUCCESS CRITERIA

What strong completion looks like

  • The workflow distinguishes analysis-ready fields from leakage-prone fields.
  • Default-risk patterns are explained with both visuals and business reasoning.
  • Evaluation choices reflect lending tradeoffs instead of relying on accuracy alone.
  • The final output makes clear what a stakeholder should trust and what still needs validation.

GUIDANCE

Support when you need a nudge

Reveal practical hints, checkpoints, and framing help without exposing a full finished solution.

Complete this project when the work is yours

Use this only after you have worked through the roadmap, milestones, and your own analysis.

KEY LEARNING OUTCOMES

Skills reinforced by this project

  • Audit a financial dataset for leakage, imbalance, and data-quality issues.
  • Use EDA and feature engineering to surface borrower-risk patterns.
  • Select classification metrics that match lending risk tradeoffs.
  • Present a modeling workflow with realistic caveats.

RELATED LESSONS

Revisit published lessons that support this build