shilpa-j-ml-package 1.0.0

A simple machine learning package built from scratch

ml_package

A custom machine learning library built from scratch — no sklearn dependency for core algorithms.

Algorithms Included

Module	Classes
linear_models	LinearRegression (OLS), Ridge (L2), Lasso (L1), ForwardSelection, BackwardElimination
decomposition	PCA
neighbors	KNNClassifier (city-block distance)
neural_network	NeuralNetwork, Perceptron
preprocessing	StandardScaler, MinMaxScaler, SimpleImputer, OutlierHandler, LabelEncoder, OneHotEncoder, Normalizer
metrics	r2_score, rmse, mae, accuracy_score, confusion_matrix, f1_score
model_selection	train_test_split
pipeline	Pipeline

Install

pip install -e .

Quick Start

import numpy as np
from ml_package.linear_models.linear_regression import LinearRegression
from ml_package.metrics.regression import r2_score

X = np.random.randn(100, 3)
y = 3*X[:,0] + 2*X[:,1] + np.random.randn(100)*0.5

model = LinearRegression()
model.fit(X, y)
print(r2_score(y, model.predict(X)))

# Check all assumptions
model.check_normality()
model.check_multicollinearity()
model.check_heteroscedasticity()
model.plot_diagnostics()

from ml_package.decomposition.pca import PCA

pca = PCA(n_components=2)
X_pca = pca.fit_transform(X)
pca.plot_components(X)

from ml_package.neighbors.knn_classifier import KNNClassifier

knn = KNNClassifier(k=3, metric="cityblock")
knn.fit(X_train, y_train)
preds = knn.predict(X_test)

from ml_package.neural_network.neural_network import NeuralNetwork
from ml_package.neural_network.perceptron import Perceptron

nn = NeuralNetwork(hidden_size=8, learning_rate=0.05, max_iter=1000)
nn.fit(X, y)

p = Perceptron(learning_rate=0.1, max_iter=200)
p.fit(X, y)

Run Tests

cd tests
python test_linear_regression.py
python test_pca.py
python test_knn.py
python test_neural_network.py

License

MIT