Machine Learning Internship Project

End-to-End Machine Learning Pipeline | Regression • Classification • Advanced Models

---

Overview

This repository showcases my internship work where I implemented machine learning models across multiple levels — from basic regression to advanced ensemble and SVM models.

The project demonstrates a complete ML workflow:

Data Preprocessing → Model Building → Evaluation → Visualization → Insights

---

Key Highlights

Implemented 3 Levels of Machine Learning Tasks
Built Regression + Classification + Advanced Models
Visualized results using graphs and confusion matrices
Performed feature importance analysis
Compared multiple models for performance

---

Repository Structure

ML-Internship-Project/ │ ├── Level1_HousePrice_Regression.ipynb ├── Churn_Prediction_Level2_Level3.ipynb │ ├── churn-bigml-80.csv (Training Data) ├── churn-bigml-20.csv (Testing Data)

---

Level 1: House Price Prediction

Goal: Predict house prices using regression

Workflow

• Data preprocessing
• Feature scaling
• Linear Regression model

Evaluation

• R² Score
• Mean Squared Error

---

Level 2: Customer Churn Prediction

Goal: Predict whether a customer will churn

Models Used

• Logistic Regression
• Decision Tree

Evaluation

• Accuracy Score
• Confusion Matrix
• Classification Report

---

Level 3: Advanced Machine Learning

Goal: Improve model performance using advanced techniques

Models Used

• Random Forest
• Support Vector Machine (SVM)

Advanced Analysis

• Feature Importance (Random Forest)
• Model Comparison
• Kernel comparison (Linear vs RBF)

---

Results & Insights

Decision Tree achieved the highest accuracy
Random Forest performed strongly with balanced predictions
Logistic Regression provided a solid baseline
SVM (Linear) struggled with class imbalance
SVM (RBF) improved performance significantly

Model Performance Comparison

Model	Accuracy
Logistic Regression	0.853
Decision Tree	0.954
Random Forest	0.945
SVM (Linear)	0.858
SVM (RBF)	0.919

Note: Accuracy alone is not sufficient due to class imbalance, so confusion matrix and recall were also analyzed.

Tech Stack

• Python
• pandas
• scikit-learn
• matplotlib & seaborn

---

How to Run

git clone https://github.com/anukta-acharya/ML-Internship-Project.git
cd ML-Internship-Project

Open notebooks in Jupyter / Colab and run cells sequentially

Conclusion

This project demonstrates a complete machine learning pipeline from basic regression to advanced classification models, focusing on performance, interpretability, and real-world insights.