In this notebook, we cover:
| Part | Topic | Time |
|---|---|---|
| 1 | Case understanding and first model | 15 min |
| 2 | Probabilities and thresholds | 10 min |
| 3 | Confusion matrix and metrics | 10 min |
| 4 | ROC and AUC | 10 min |
| 5 | Comparing customer categories | 10 min |
| Wrap-up | Reflection | 5 min |
You are a business analyst working for LendWise, a fintech company that offers fast digital consumer loans. Over the last six months, LendWise has seen a noticeable spike in credit defaults in two customer categories:
Early-career urban borrowers: Typically younger customers, often with lower savings, moderate income, and irregular repayment histories.
Self-employed platform workers: Customers with fluctuating income, often higher short-term debt exposure, and less predictable monthly cash flows.
The risk team wants to understand these groups better. They are especially interested in whether default risk follows the same logic in both categories, or whether the predictors and model quality differ substantially.
In this notebook, you first focus on early-career urban borrowers and build a logistic regression model to predict default. Afterward, you compare the results with self-employed platform workers.
LendWise is interested in questions such as:
The dataset contains the following variables:
customer_category: customer segmentincome: monthly income in EURdebt_to_income: debt-to-income ratiomissed_payments_12m: number of missed payments in the last 12 monthscredit_utilization: share of available credit currently used (between 0 and 1)months_with_company: months since first activity with LendWisedefault: whether the customer defaulted (1 = yes, 0 = no)Note: The dataset was curated by the company’s Risk Analytics Department. It has already been cleaned, validated, and prepared for modeling.
import pandas as pd
df = pd.read_csv("data/credit_default.csv")
df.head()| customer_category | income | debt_to_income | missed_payments_12m | credit_utilization | months_with_company | default | |
|---|---|---|---|---|---|---|---|
| 0 | early_career_urban | 3283.287284 | 0.453151 | 2.0 | 0.574281 | 21.0 | 0 |
| 1 | platform_worker | 1976.088070 | 0.597144 | 0.0 | 0.855863 | 39.0 | 1 |
| 2 | platform_worker | 3299.177450 | 0.664904 | 4.0 | 0.969859 | 27.0 | 1 |
| 3 | platform_worker | 2539.814969 | 0.505792 | 0.0 | 0.939466 | 7.0 | 0 |
| 4 | early_career_urban | 1655.521735 | 0.342233 | 2.0 | 0.461253 | 30.0 | 1 |
In this first part, focus only on early-career urban borrowers.
Create a filtered dataset that contains only customers from early-career urban borrowers.
# Your solution here
# df_ec_urban = ...Estimate a logistic regression model to predict credit default.
default)from sklearn.linear_model import LogisticRegression
# Your solution here...Interpret the coefficients in substantive terms:
Explain how the model moves from:
In this part, we move from model coefficients to predicted probabilities and classification decisions.
Logistic regression outputs probabilities, which must be converted into classes using a threshold. The threshold directly affects business outcomes: lowering it increases approvals (but also defaults), while raising it reduces risk (but may reject good customers).
Run the code and answer the following questions:
y_prob = model.predict_proba(X)[:, 1]
y_pred = model.predict(X)
y_prob[:5], y_pred[:5]y_prob and y_pred?predict_proba return two columns?Instead of relying on the default threshold of 0.5, we can manually adjust the cutoff to reflect different business objectives. In practice, thresholds are used to identify specific customer groups, for example:
In general, predictions can be obtained for any threshold t as:
t = 0.3
y_pred_t = (y_prob >= t).astype(int)
y_pred_tFirst, explore different thresholds by modifying t (e.g., 0.2, 0.4, 0.6, 0.7):
Now consider a concrete business task: LendWise wants to identify the 20% most risky customers to prioritize early intervention.
# Your solution here ...
# Hint: use the `np.quantile()` function on the predicted probabilitiesSimilarly, LendWise may want to identify the 20% safest customers:
# Your solution here ...For simplicity, model performance is evaluated on the same data used for estimation. In practice, a train/test split or cross-validation should be used.
Use the confusion_matrix function of sklearn.metrics to generate the confusion matrix:
from sklearn.metrics import confusion_matrix
cm = confusion_matrix(y, y_pred)
cmThe confusion_matrix function returns an unlabeled ndarray. Consult with the documentation to identify:
Compute and interpret:
LendWise is currently trying to grow its customer base aggressively.
Discuss:
Run the following code to compute and visualize the ROC curve:
from sklearn.metrics import roc_curve, auc
import matplotlib.pyplot as plt
fpr, tpr, thresholds = roc_curve(y, y_prob)
roc_auc = auc(fpr, tpr)
# Plot ROC curve
plt.figure()
plt.plot(fpr, tpr, label=f"AUC = {roc_auc:.2f}")
plt.plot([0, 1], [0, 1], linestyle="--") # random baseline
plt.xlabel("False Positive Rate")
plt.ylabel("True Positive Rate")
plt.title("ROC Curve")
plt.legend()
plt.show()Now repeat the same modeling steps for the segment of platform workers.
You do not need to interpret every single line of code again. Focus on comparing the results.
# Your code here ...Fit a logistic regression model for platform workers.
# Your code here ...Compare the coefficients across both categories.
# Your code here ...Compare the ROC curves and AUC values of the two models.
# Your code here ...🎉🎈 You have completed the notebook – good work! 🎈🎉
In this notebook, we have learned to:
scikit-learnBefore you wrap up, please complete the Session 5 survey here: ?meta:surveys.session_05.url. Thank you 🙏