🧠 AI with Python – 🏦 Loan Approval Prediction

Loan approval is one of the most common and impactful decision-making problems in banking and fintech.

Every loan approval involves balancing business growth with financial risk, making accurate and explainable predictions essential.

In this project, we build a loan approval prediction model using supervised machine learning on tabular financial data, demonstrating how ML assists real-world credit decisioning systems.

---

Understanding the Problem

Loan datasets typically contain information about applicants such as income, credit history, employment status, and loan amount.

The challenge lies in:

• handling missing and categorical data
• managing class imbalance
• minimizing risky approvals (false positives)
• ensuring interpretability for regulatory compliance

Unlike academic problems, accuracy alone is not sufficient in financial decision systems.

---

1. Loading the Loan Dataset

We start with a loan approval dataset in CSV format.

import pandas as pd

df = pd.read_csv("loan_data.csv")
df.head()

The target variable indicates whether a loan was approved or rejected.

---

2. Inspecting Data Quality

Before modeling, we examine data structure and class distribution.

print(df.info())
print(df["loan_status"].value_counts())

Loan datasets often contain missing values and mixed data types, which must be addressed carefully.

---

3. Handling Missing Values

Missing data is common in real-world financial records.

We use simple but effective imputation strategies.

from sklearn.impute import SimpleImputer

num_cols = df.select_dtypes(include="number").columns
cat_cols = df.select_dtypes(exclude="number").columns.drop("loan_status")

num_imputer = SimpleImputer(strategy="median")
cat_imputer = SimpleImputer(strategy="most_frequent")

df[num_cols] = num_imputer.fit_transform(df[num_cols])
df[cat_cols] = cat_imputer.fit_transform(df[cat_cols])

This ensures the dataset is complete and model-ready.

---

4. Encoding Categorical Features

Machine learning models require numeric inputs, so categorical fields must be encoded.

from sklearn.preprocessing import LabelEncoder

encoder = LabelEncoder()
for col in cat_cols:
    df[col] = encoder.fit_transform(df[col])

This step converts categories like employment type or marital status into numeric form.

---

5. Train/Test Split with Stratification

We split the dataset while preserving class proportions.

from sklearn.model_selection import train_test_split

X = df.drop("loan_status", axis=1)
y = df["loan_status"]

X_train, X_test, y_train, y_test = train_test_split(
    X, y,
    test_size=0.3,
    stratify=y,
    random_state=42
)

Stratification ensures fair evaluation across approval and rejection cases.

---

6. Training a Loan Approval Model

We use Logistic Regression as a baseline model due to its interpretability and stability.

from sklearn.linear_model import LogisticRegression

model = LogisticRegression(max_iter=1000)
model.fit(X_train, y_train)

Logistic Regression provides probability outputs that align well with financial risk scoring.

---

7. Evaluating Model Performance

Evaluation focuses on metrics relevant to financial risk.

from sklearn.metrics import classification_report, confusion_matrix

y_pred = model.predict(X_test)

print(classification_report(y_test, y_pred))
print(confusion_matrix(y_test, y_pred))

Precision is especially important to reduce the number of risky loan approvals.

---

8. ROC–AUC for Approval Ranking

ROC–AUC measures how well the model ranks approved vs rejected applicants.

from sklearn.metrics import roc_auc_score

y_proba = model.predict_proba(X_test)[:, 1]
auc = roc_auc_score(y_test, y_proba)

print("ROC–AUC:", auc)

This metric is threshold-independent and widely used in credit modeling.

---

Key Takeaways

1. Loan approval prediction is a core real-world ML use case in finance.
2. Data preprocessing plays a critical role in model reliability.
3. Logistic Regression provides interpretable and stable risk scores.
4. Precision and ROC–AUC are more meaningful than accuracy in credit decisions.
5. ML assists decision-making but does not replace business rules or compliance checks.

---

Conclusion

Loan approval prediction showcases how machine learning supports high-stakes financial decisions.

By combining careful preprocessing, interpretable models, and appropriate evaluation metrics, ML systems can help institutions make faster, more consistent, and data-driven credit decisions.

This project represents a realistic end-to-end fintech ML workflow, making it a strong addition to the AI with Python – Real-World Mini Projects (Advanced) series.

---

import pandas as pd

df = pd.read_csv("loan_data.csv")
df.head()


print(df.info())
print(df["loan_status"].value_counts())


from sklearn.impute import SimpleImputer

num_cols = df.select_dtypes(include="number").columns
cat_cols = df.select_dtypes(exclude="number").columns.drop("loan_status")

num_imputer = SimpleImputer(strategy="median")
cat_imputer = SimpleImputer(strategy="most_frequent")

df[num_cols] = num_imputer.fit_transform(df[num_cols])
df[cat_cols] = cat_imputer.fit_transform(df[cat_cols])


from sklearn.preprocessing import LabelEncoder

encoder = LabelEncoder()
for col in cat_cols:
    df[col] = encoder.fit_transform(df[col])


from sklearn.model_selection import train_test_split

X = df.drop("loan_status", axis=1)
y = df["loan_status"]

X_train, X_test, y_train, y_test = train_test_split(
    X, y,
    test_size=0.3,
    stratify=y,
    random_state=42
)

from sklearn.linear_model import LogisticRegression

model = LogisticRegression(max_iter=1000)
model.fit(X_train, y_train)


from sklearn.metrics import classification_report, confusion_matrix

y_pred = model.predict(X_test)

print(classification_report(y_test, y_pred))
print(confusion_matrix(y_test, y_pred))


from sklearn.metrics import roc_auc_score

y_proba = model.predict_proba(X_test)[:, 1]
auc = roc_auc_score(y_test, y_proba)

print("ROC–AUC:", auc)