Modeling Car Insurance Claim Outcomes

This project builds a predictive model for On the Road car insurance to identify which customers are likely to make a claim during their policy period. Given the company's limited machine learning infrastructure, the goal is to find the single most predictive feature that delivers the highest model accuracy. The analysis uses logistic regression on a cleaned customer dataset to determine the optimal simple model for potential deployment.

This project was completed using DataCamp’s Datalab environment.

---

🎯 Project Objectives

• Load and explore customer data related to car insurance claims
• Handle missing values in key variables
• Train individual logistic regression models using each feature separately
• Evaluate model performance by accuracy, calculated using a confusion matrix
• Identify the single best-performing feature for predicting insurance claims

---

🗃️ Dataset Overview

The data comes from a single CSV file:

File	Description
car_insurance.csv	Customer profiles and claim history

Final Dataset Columns

Column	Description
id	Unique client identifier
age	Client's age group: 0 (16–25), 1 (26–39), 2 (40–64), 3 (65+)
gender	Client's gender: 0 (Female), 1 (Male)
driving_experience	Years with a license: 0 (0–9), 1 (10–19), 2 (20–29), 3 (30+)
education	Level of education: 0 (No education), 1 (High school), 2 (University)
income	Income level: 0 (Poverty), 1 (Working class), 2 (Middle class), 3 (Upper class)
credit_score	Credit score (continuous, 0 to 1)
vehicle_ownership	Ownership status: 0 (Financing), 1 (Owns vehicle)
vehicle_year	Vehicle registration year: 0 (Before 2015), 1 (2015 or later)
married	Marital status: 0 (Not married), 1 (Married)
children	Number of children
postal_code	Client's postal code
annual_mileage	Annual miles driven (numeric)
vehicle_type	Type of car: 0 (Sedan), 1 (Sports car)
speeding_violations	Number of speeding tickets
duis	Number of DUI offenses
past_accidents	Number of previous accidents
outcome	Target variable: 0 (No claim), 1 (Made claim)

---

🔍 Key Findings

• ✅ Best predictive feature: driving_experience
• 📊 Highest model accuracy: 77.71%
• 🧠 Among all individual features, driving experience was the strongest predictor of whether a customer would make a claim
• 🛠️ Missing values in credit_score (9.82%) and annual_mileage (9.57%) were imputed using the median value

---

🛠️ Tools Used

• Python
• pandas for data loading and preprocessing
• statsmodels (logit) for logistic regression modeling
• NumPy for numerical operations
• Jupyter Notebook / DataLab for analysis and visualization

---

📌 How to Use

1. Clone or download this repository
2. Place the car_insurance.csv file in the /data folder
3. Open the notebook Modeling_Car_Insurance_Claim_Outcomes.ipynb in Jupyter or any compatible environment
4. Run the cells to reproduce the analysis
5. Modify the modeling approach to test combinations of features or alternative algorithms (e.g., decision trees)

---

✍️ Author

Project by Achraf Salimi — part of an ongoing journey to build and showcase data science skills for real-world business impact.