Machine Learning From Scratch in Python

This project implements core Machine Learning algorithms entirely from scratch using only Python and NumPy, without using scikit-learn for model training.

It is built to demonstrate:

• Strong understanding of ML fundamentals
• Research-level implementation skills
• Internship and placement readiness
• A high-quality GitHub portfolio project

---

Implemented Algorithms

Algorithm	Type	Implemented From Scratch
Linear Regression	Regression	Yes
Logistic Regression	Classification	Yes
K-Nearest Neighbors (KNN)	Classification & Regression	Yes
Decision Tree (CART)	Classification	Yes
Random Forest	Ensemble Learning	Yes

---

Key Features

• Custom Train-Test Split implementation
• Custom Evaluation Metrics
• Full Machine Learning Pipeline System
• Custom Standard Scaler
• No dependency on scikit-learn for training
• Data visualization using Matplotlib
• Clean, modular, and scalable project structure

---

Project Structure

ml-from-scratch/ │ ├── data/ │ ├── linear_data.csv │ ├── logistic_data.csv │ └── tree_data.csv │ ├── models/ │ ├── linear_regression.py │ ├── logistic_regression.py │ ├── knn.py │ ├── decision_tree.py │ └── random_forest.py │ ├── utils/ │ ├── train_test_split.py │ ├── metrics.py │ ├── scaler.py │ └── pipeline.py │ ├── main.py ├── run_knn.py ├── run_tree.py ├── run_random_forest.py ├── run_pipeline.py └── README.md

---

How to Run the Project

Activate virtual environment:

venv\Scripts\activate

Run different models:
python main.py
python run_knn.py
python run_tree.py
python run_random_forest.py
python run_pipeline.py


This repository demonstrates core machine learning engineering skills, focusing on mathematical correctness, algorithmic clarity, and clean system design, rather than simply using pre-built libraries.