Analysing Bank Loan Approval Data

import pandas as pd
import numpy as np

# Build the CrestBank loan applications dataset
df = pd.DataFrame({
    "app_id":          ["A001","A002","A003","A004","A005","A006",
                        "A007","A008","A009","A010","A011","A012"],
    "annual_income":   [72000,34000,95000,41000,58000,120000,
                        29000,83000,47000,67000,38000,110000],
    "credit_score":    [745,612,790,638,710,810,
                        590,755,665,720,625,780],
    "loan_amount":     [18000,22000,35000,28000,15000,40000,
                        20000,25000,30000,22000,18000,45000],
    "loan_term_yrs":   [5,7,10,7,5,10,7,5,10,5,7,10],
    "employment_type": ["Salaried","Self-Employed","Salaried","Contract","Salaried",
                        "Salaried","Self-Employed","Salaried","Contract","Salaried",
                        "Self-Employed","Salaried"],
    "approved":        ["Yes","No","Yes","No","Yes","Yes",
                        "No","Yes","No","Yes","No","Yes"]
})

# Basic inspection
print("Shape:", df.shape)
print("Missing values:", df.isnull().sum().sum())

# Create binary approval flag — needed for all numeric calculations
df["approved_flag"] = (df["approved"] == "Yes").astype(int)

# Overall approval rate
overall_rate = df["approved_flag"].mean()
print(f"\nOverall approval rate: {overall_rate:.1%}")
print(f"Approved: {df['approved_flag'].sum()} | Rejected: {(df['approved_flag']==0).sum()}")

# Approval rate by employment type
emp_approval = df.groupby("employment_type")["approved_flag"].agg(
    count="count",
    approved="sum",
    approval_rate="mean"
).round(2).reset_index()
emp_approval["approval_rate"] = (emp_approval["approval_rate"] * 100).round(0)

print("\nApproval rate by employment type:")
print(emp_approval.to_string(index=False))

# Profile comparison — mean values for approved vs rejected applicants
profile = df.groupby("approved")[
    ["annual_income","credit_score","loan_amount","loan_term_yrs"]
].mean().round(0)

print("Average profile by approval outcome:")
print(profile.to_string())

# Compute the gap between approved and rejected on each variable
gap = profile.loc["Yes"] - profile.loc["No"]
print("\nGap (Approved minus Rejected):")
print(gap.to_string())

# Express income gap as a percentage above rejected applicants
income_pct = (gap["annual_income"] / profile.loc["No","annual_income"] * 100)
score_pct  = (gap["credit_score"]  / profile.loc["No","credit_score"]  * 100)
print(f"\nApproved applicants earn {income_pct:.1f}% more than rejected applicants")
print(f"Approved applicants have credit scores {score_pct:.1f}% higher than rejected")

# Compute debt-to-income ratio: loan amount as a % of annual income
df["dti_ratio"] = (df["loan_amount"] / df["annual_income"] * 100).round(1)

# Show all applicants with their DTI and approval outcome
dti_view = df[["app_id","annual_income","loan_amount","dti_ratio","approved"]].sort_values(
    "dti_ratio", ascending=False
)
print("All applicants sorted by DTI ratio:")
print(dti_view.to_string(index=False))

# Mean DTI for approved vs rejected
dti_by_outcome = df.groupby("approved")["dti_ratio"].agg(["mean","max","min"]).round(1)
print("\nDTI statistics by approval outcome:")
print(dti_by_outcome.to_string())

# Flag high-DTI applicants (above 40%) and check their outcomes
high_dti = df[df["dti_ratio"] > 40][["app_id","dti_ratio","approved","credit_score"]]
print(f"\nHigh DTI applicants (>40%): {len(high_dti)}")
print(high_dti.to_string(index=False))

# Credit score threshold analysis
min_approved_score = df[df["approved"] == "Yes"]["credit_score"].min()
max_rejected_score = df[df["approved"] == "No"]["credit_score"].max()

print(f"Minimum credit score among APPROVED applicants: {min_approved_score}")
print(f"Maximum credit score among REJECTED applicants: {max_rejected_score}")
print(f"Overlap zone: {max_rejected_score} to {min_approved_score}")

# Applicants in the overlap zone (if any)
overlap = df[
    (df["credit_score"] >= max_rejected_score) &
    (df["credit_score"] <= min_approved_score)
][["app_id","credit_score","annual_income","dti_ratio","approved"]]
print(f"\nApplicants in the credit score overlap zone: {len(overlap)}")
if len(overlap) > 0:
    print(overlap.to_string(index=False))

# Correlation of each numeric variable with approval flag
corr_vars = ["annual_income","credit_score","loan_amount","loan_term_yrs","dti_ratio"]
print("\nCorrelation with approval (Pearson r):")
for var in corr_vars:
    r = df[var].corr(df["approved_flag"]).round(3)
    print(f"  {var:<18} r = {r:+.3f}")