Analysing Credit Risk

import pandas as pd
import numpy as np

df = pd.DataFrame({
    "app_id":           ["A001","A002","A003","A004","A005","A006","A007","A008",
                         "A009","A010","A011","A012","A013","A014","A015","A016",
                         "A017","A018","A019","A020"],
    "income_annual":    [62000,34000,88000,41000,75000,29000,110000,52000,
                         38000,68000,45000,92000,31000,58000,84000,36000,
                         72000,48000,27000,95000],
    "employment":       ["Employed","Self-Employed","Employed","Contract","Employed",
                         "Self-Employed","Employed","Contract","Self-Employed","Employed",
                         "Contract","Employed","Self-Employed","Employed","Employed",
                         "Contract","Employed","Self-Employed","Self-Employed","Employed"],
    "loan_amount":      [18000,22000,45000,28000,32000,18500,55000,31000,
                         24000,28000,29000,48000,20000,22000,42000,21000,
                         35000,26000,17000,50000],
    "collateral_value": [28000,24000,72000,30000,55000,20000,90000,34000,
                         25000,48000,32000,78000,21000,35000,68000,22000,
                         58000,28000,18000,85000],
    "monthly_debt":     [820,1240,1480,1620,1100,1380,1920,1740,
                         1560,980,1680,1540,1420,860,1260,1580,
                         1140,1480,1320,1640],
    "credit_score":     [718,634,762,601,741,588,798,622,
                         612,734,598,771,574,722,756,595,
                         728,641,562,784],
    "defaulted":        [0,1,0,1,0,1,0,1,
                         1,0,1,0,1,0,0,1,
                         0,1,1,0]
})

print(f"Portfolio: {len(df)} loans | Defaults: {df['defaulted'].sum()} | Default rate: {df['defaulted'].mean():.1%}")

# Derived metrics
df["monthly_income"] = df["income_annual"] / 12
df["dti"]            = (df["monthly_debt"] / df["monthly_income"] * 100).round(1)
df["ltv"]            = (df["loan_amount"] / df["collateral_value"] * 100).round(1)

# Risk flags
DTI_THRESHOLD = 43.0   # % — standard affordability threshold
LTV_THRESHOLD = 90.0   # % — high LTV threshold

df["high_dti"] = (df["dti"] > DTI_THRESHOLD).astype(int)
df["high_ltv"] = (df["ltv"] > LTV_THRESHOLD).astype(int)

print("\nDTI and LTV distribution:")
print(df[["dti","ltv"]].describe().round(1))

print("\nApplications breaching DTI > 43%:", df["high_dti"].sum())
print("Applications breaching LTV > 90%:", df["high_ltv"].sum())

# Show the derived risk profile
print("\nFull risk profile:")
print(df[["app_id","income_annual","employment","dti","ltv",
          "credit_score","defaulted"]].to_string(index=False))

# Income bands using pd.cut()
df["income_band"] = pd.cut(
    df["income_annual"],
    bins   = [0, 40000, 65000, 90000, float("inf")],
    labels = ["Low (<40k)","Mid (40–65k)","High (65–90k)","Premium (90k+)"]
)

# DTI tiers
df["dti_tier"] = pd.cut(
    df["dti"],
    bins   = [0, 36, 43, 55, float("inf")],
    labels = ["Low","Moderate","High","Very High"]
)

# Default rate by employment type
emp_default = df.groupby("employment").agg(
    applications  = ("app_id",    "count"),
    defaults      = ("defaulted", "sum"),
    default_rate  = ("defaulted", "mean"),
    mean_dti      = ("dti",       "mean"),
    mean_score    = ("credit_score","mean")
).round(2).reset_index()
emp_default["rate_pct"] = (emp_default["default_rate"] * 100).round(0)
emp_default = emp_default.sort_values("default_rate", ascending=False)

print("Default rate by employment type:")
print(emp_default[["employment","applications","defaults",
                    "rate_pct","mean_dti","mean_score"]].to_string(index=False))

# Default rate by income band
inc_default = df.groupby("income_band", observed=True).agg(
    applications = ("app_id",    "count"),
    defaults     = ("defaulted", "sum"),
    default_rate = ("defaulted", "mean"),
    mean_dti     = ("dti",       "mean")
).round(2).reset_index()
inc_default["rate_pct"] = (inc_default["default_rate"] * 100).round(0)

print("\nDefault rate by income band:")
print(inc_default[["income_band","applications","defaults",
                    "rate_pct","mean_dti"]].to_string(index=False))

# Pivot: default rate by employment × income band
pivot_default = pd.pivot_table(
    df,
    index   = "employment",
    columns = "income_band",
    values  = "defaulted",
    aggfunc = "mean",
    fill_value = 0
).round(2)

print("Default rate pivot (employment × income band):")
print((pivot_default * 100).round(0).to_string())

# Credit score band analysis
df["score_band"] = pd.cut(
    df["credit_score"],
    bins   = [0, 599, 649, 699, 749, 850],
    labels = ["<600","600–649","650–699","700–749","750+"]
)

score_default = df.groupby("score_band", observed=True).agg(
    count        = ("app_id",     "count"),
    defaults     = ("defaulted",  "sum"),
    default_rate = ("defaulted",  "mean"),
    mean_dti     = ("dti",        "mean")
).round(2).reset_index()
score_default["rate_pct"] = (score_default["default_rate"] * 100).round(0)

print("\nDefault rate by credit score band:")
print(score_default[["score_band","count","defaults",
                      "rate_pct","mean_dti"]].to_string(index=False))

# Credit score correlation with default
score_corr = df["credit_score"].corr(df["defaulted"]).round(3)
dti_corr   = df["dti"].corr(df["defaulted"]).round(3)
ltv_corr   = df["ltv"].corr(df["defaulted"]).round(3)
print(f"\nCorrelations with default:")
print(f"  credit_score: {score_corr:+.3f}")
print(f"  dti:          {dti_corr:+.3f}")
print(f"  ltv:          {ltv_corr:+.3f}")

# Full profile comparison: defaulted vs performing
profile = df.groupby("defaulted").agg(
    count         = ("app_id",        "count"),
    mean_income   = ("income_annual", "mean"),
    mean_dti      = ("dti",           "mean"),
    mean_ltv      = ("ltv",           "mean"),
    mean_score    = ("credit_score",  "mean"),
    pct_employed  = ("employment",    lambda x: (x == "Employed").mean() * 100)
).round(1)
profile.index = profile.index.map({0:"Performing", 1:"Defaulted"})

print("Borrower profile: Performing vs Defaulted:")
print(profile.to_string())

# Identify the single riskiest applicant profile
# High DTI + low score + non-employed
df["risk_score"] = (
    (df["dti"]          > DTI_THRESHOLD).astype(int) +
    (df["credit_score"] < 650           ).astype(int) +
    (df["employment"]  != "Employed"    ).astype(int) +
    (df["ltv"]          > LTV_THRESHOLD ).astype(int)
)

print("\nComposite risk score distribution:")
print(df.groupby("risk_score")["defaulted"].agg(
    count="count", defaults="sum",
    default_rate="mean"
).round(2).to_string())

# Proposed policy recommendation summary
print("\nProposed credit policy tightening:")
approved_new = df[
    (df["credit_score"] >= 700) &
    (df["employment"]   == "Employed") &
    (df["dti"]          <= 43)
]
print(f"  Applicants passing new criteria: {len(approved_new)} of {len(df)}")
print(f"  Default rate among approved:     {approved_new['defaulted'].mean():.1%}")
print(f"  Defaults avoided vs current:     {df['defaulted'].sum() - approved_new['defaulted'].sum()}")