Customer-Insurance-Purchase-Prediction-Using-Machine-Learning-Classification-Algorithms

🛡️ Customer Insurance Purchases Prediction - ML Case Study

📌 Project Overview

In this project, we act as analysts for a Bank Insurance Company. The main goal is to build machine learning models that can predict whether a new customer will purchase insurance based on their Age and Estimated Salary.

We perform a comparative analysis using multiple ML classification algorithms to identify the most effective model that balances both accuracy and generalization (avoiding overfitting).

---

📊 Problem Statement

Use customer data (excluding personal information) to train machine learning classifiers and evaluate their performance. The objective is to select the best-performing model based on metrics such as accuracy, precision, recall, F1-score, and ROC AUC.

---

🔍 Dataset

• Source: Social_Network_Ads.csv
• Features used:
  • Age
  • EstimatedSalary
• Target:
  • Purchased (1 = Purchased Insurance, 0 = Not Purchased)

---

⚙️ Machine Learning Algorithms Used

1. Logistic Regression
2. K-Nearest Neighbors (KNN)
3. Support Vector Machine (SVM)
4. Decision Tree
5. Random Forest

---

🧪 Steps Followed

1. Data Preprocessing
   • Load and explore dataset
   • Feature scaling (Standardization)
   • Train-test split (80-20)
2. Model Training
   • Fit each algorithm on training data
3. Evaluation Metrics
   • Accuracy
   • Confusion Matrix
   • Classification Report
   • ROC AUC Curve
4. Prediction & Visualization
   • Predict insurance purchase for custom Age & Salary values
   • Graphical analysis of decision boundaries

---

🎯 Graphical Predictions

We visualize model decision boundaries and test predictions for:

• Age: 30, Salary: 87,000
• Age: 40, Salary: None
• Age: 40, Salary: 100,000
• Age: 50, Salary: None
• Age: 18, Salary: None
• Age: 22, Salary: 600,000
• Age: 35, Salary: 2,500,000
• Age: 60, Salary: 100,000,000

---

📌 Hypotheses & Assumptions Tested

• Younger individuals with high salaries are more likely to purchase insurance.
• Salary may influence insurance purchase decisions more than age.
• Older users with high income might still avoid buying insurance.

---

📈 Final Output

• Tabular comparison of model performance
• Plots showing decision boundaries and prediction zones
• A final chosen model for deployment
• Code to predict on new data
• Assumptions validated through AI simulations

---

💡 Lessons Learned

• Logistic Regression and Random Forest are effective for binary classification.
• Importance of scaling for algorithms like KNN and SVM.
• Visualizations help in understanding model behavior clearly.

---

🚀 Real-life Use Cases

1. Marketing Campaigns for Insurance

   • Target customers more likely to buy insurance based on age and income.

2. Bank Loan Approvals

   • Predict customer creditworthiness using similar classification strategies.

---

🛠️ Tech Stack

• Python 3.x
• Jupyter Notebook
• Libraries:
  • pandas, numpy
  • matplotlib, seaborn
  • scikit-learn

---