Bhagya My ML Package

A lightweight Machine Learning package built from scratch in Python for educational purposes and algorithm understanding.

This package implements core Machine Learning algorithms without relying on scikit-learn model implementations, making it useful for students who want to learn how ML algorithms work internally.

Features

Linear Models

• Linear Regression
• Intercept and coefficient calculation
• Prediction using Normal Equation and Pseudoinverse

Nearest Neighbors

• KNN Classifier
• KNN Regressor
• Euclidean Distance based similarity

Neural Networks

• Feed Forward Neural Network
• Backpropagation Learning
• Sigmoid Activation Function
• Hebbian Learning Network

Dimensionality Reduction

• Principal Component Analysis (PCA)

Preprocessing

• StandardScaler
• Feature Standardization

Model Selection

• Train-Test Split

Metrics

Regression Metrics

• R² Score
• Adjusted R² Score
• Mean Squared Error (MSE)
• Mean Absolute Error (MAE)
• Root Mean Squared Error (RMSE)

Classification Metrics

• Accuracy Score

Installation

pip install bhagya-my-ml-package

Example Usage

Linear Regression

from my_ml_package.linear_models import LinearRegression

model = LinearRegression()

model.fit(X_train, y_train)

predictions = model.predict(X_test)

print(model.intercept_)
print(model.coef_)

KNN Regression

from my_ml_package.neighbors import KNNRegressor

model = KNNRegressor(k=3)

model.fit(X_train, y_train)

predictions = model.predict(X_test)

PCA

from my_ml_package.decomposition import PCA

pca = PCA(n_components=2)

X_reduced = pca.fit_transform(X)

StandardScaler

from my_ml_package.preprocessing import StandardScaler

scaler = StandardScaler()

X_scaled = scaler.fit_transform(X)

Algorithms Implemented

• Linear Regression
• KNN Classification
• KNN Regression
• Artificial Neural Network
• Hebbian Learning Network
• Principal Component Analysis (PCA)

Educational Objective

The primary goal of this package is to help students understand Machine Learning algorithms by implementing them from scratch using NumPy and core Python concepts.

Author

Bhagyashree Patil

License

MIT License