Car_Insurance.py

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

from sklearn.model_selection import train_test_split
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import classification_report, confusion_matrix

# -------------------------
# Part 1: Load the Dataset
# -------------------------
df = pd.read_csv('Car_Insurance_Claim.csv')
print("Data Shape:", df.shape)

# -----------------------------
# Part 2: Basic Data Exploration
# -----------------------------
df.info()
print("\nDescriptive Statistics:")
print(df.describe(include='all'))

# -------------------------------
# Part 3: Data Cleaning & Prep
# -------------------------------
# 3.1 Check for duplicates
duplicate_rows = df.duplicated().sum()
print("\nNumber of duplicate rows:", duplicate_rows)

# 3.2 Check for missing values
missing_values = df.isnull().sum()
print("\nMissing values:\n", missing_values)

# Convert relevant columns to categorical
categorical_cols = ['AGE','GENDER','RACE','DRIVING_EXPERIENCE',
                    'EDUCATION','INCOME','VEHICLE_TYPE']
for col in categorical_cols:
    df[col] = df[col].astype('category')

# Convert binary flags to int
df['VEHICLE_OWNERSHIP'] = df['VEHICLE_OWNERSHIP'].astype(int)
df['MARRIED'] = df['MARRIED'].astype(int)
df['CHILDREN'] = df['CHILDREN'].astype(int)

# Ensure numeric columns are indeed numeric
numeric_cols = ['CREDIT_SCORE','ANNUAL_MILEAGE','SPEEDING_VIOLATIONS',
                'DUIS','PAST_ACCIDENTS','OUTCOME']
for col in numeric_cols:
    df[col] = pd.to_numeric(df[col], errors='coerce')

# 3.3 Impute missing values with the mean
df['CREDIT_SCORE'] = df['CREDIT_SCORE'].fillna(df['CREDIT_SCORE'].mean())
df['ANNUAL_MILEAGE'] = df['ANNUAL_MILEAGE'].fillna(df['ANNUAL_MILEAGE'].mean())

df.info()  # Final check after cleaning

# -------------------------------------
# Part 4: Exploratory Data Analysis (EDA)
# -------------------------------------
# 4.1 Distribution of Key Numerical Variables
numeric_features = ['CREDIT_SCORE','ANNUAL_MILEAGE','SPEEDING_VIOLATIONS','DUIS','PAST_ACCIDENTS']
df[numeric_features].hist(figsize=(12, 8), bins=15, edgecolor='black')
plt.tight_layout()
plt.show()

# 4.2 Distribution of AGE Categories
plt.figure(figsize=(8,5))
sns.countplot(x='AGE', data=df, order=df['AGE'].value_counts().index)
plt.title('Distribution of AGE Categories')
plt.show()

# 4.3 Relationship Between AGE and OUTCOME
age_outcome = df.groupby('AGE')['OUTCOME'].mean().reset_index()
plt.figure(figsize=(8,5))
sns.barplot(x='AGE', y='OUTCOME', data=age_outcome, order=age_outcome['AGE'])
plt.title('Mean OUTCOME by AGE Category')
plt.xlabel('AGE Category')
plt.ylabel('Mean OUTCOME (Purchase/Claim Rate)')
plt.show()

# -----------------------------------------
# Part 5: Correlation and Pivot Table
# -----------------------------------------
plt.figure(figsize=(10,6))
corr_matrix = df[numeric_cols].corr()
sns.heatmap(corr_matrix, annot=True, cmap='Blues')
plt.title('Correlation Heatmap')
plt.show()

pivot_table_example = df.pivot_table(values='OUTCOME', index='AGE', columns='GENDER', aggfunc='mean').fillna(0)
print("\nPivot Table - Mean OUTCOME by AGE & GENDER:\n", pivot_table_example)

# -----------------------------------------------
# Part 6: Basic Predictive Modeling (Logistic Reg)
# -----------------------------------------------
df_encoded = pd.get_dummies(df, 
    columns=['AGE','GENDER','RACE','DRIVING_EXPERIENCE','EDUCATION','INCOME','VEHICLE_TYPE'], 
    drop_first=True
)

X = df_encoded.drop(['ID','POSTAL_CODE','OUTCOME','VEHICLE_YEAR'], axis=1, errors='ignore')
y = df_encoded['OUTCOME']

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)

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

y_pred = logreg.predict(X_test)

print("\nClassification Report:\n", classification_report(y_test, y_pred))
print("Confusion Matrix:\n", confusion_matrix(y_test, y_pred))

# -----------------------------------------
# Part 7: Conclusions & Recommendations
# -----------------------------------------
print("\nKey Observations:")
print("- Younger drivers (16-25) show a higher average OUTCOME rate.")
print("- Credit scores, speeding violations, DUIs, and past accidents all influence OUTCOME.")
print("- The pivot table indicates that male drivers in younger age brackets have higher OUTCOME rates.")

print("\nRecommendations:")
print("- Consider tailoring marketing or pricing strategies for younger age groups.")
print("- Further data on online interactions (e.g., time spent, quote frequency) would deepen the analysis.")