🧠 MLfromScratch

MLfromScratch is a library designed to help you learn and understand machine learning algorithms by building them from scratch using only NumPy! No black-box libraries, no hidden magic—just pure Python and math. It's perfect for beginners who want to see what's happening behind the scenes of popular machine learning models.

🔗 Explore the Documentation

---

📦 Package Structure

Our package structure is designed to look like scikit-learn, so if you're familiar with that, you'll feel right at home!

🔧 Modules and Algorithms (Explained for Beginners)

📈 1. Linear Models (linear_model)

• LinearRegression : Imagine drawing a straight line through a set of points to predict future values. Linear Regression helps in predicting something like house prices based on size.

• SGDRegressor : A fast way to do Linear Regression using Stochastic Gradient Descent. Perfect for large datasets.

• SGDClassifier : A classification algorithm predicting categories like "spam" or "not spam."
  
  

🌳 2. Decision Trees (tree)

• DecisionTreeClassifier : Think of this as playing 20 questions to guess something. A decision tree asks yes/no questions to classify data.

• DecisionTreeRegressor : Predicts a continuous number (like temperature tomorrow) based on input features.
  
  

👥 3. K-Nearest Neighbors (neighbors)

• KNeighborsClassifier : Classifies data by looking at the 'k' nearest neighbors to the new point.

• KNeighborsRegressor : Instead of classifying, it predicts a number based on nearby data points.
  
  

🧮 4. Naive Bayes (naive_bayes)

• GaussianNB : Works great for data that follows a normal distribution (bell-shaped curve).

• MultinomialNB : Ideal for text classification tasks like spam detection.
  
  

📊 5. Clustering (cluster)

• KMeans : Groups data into 'k' clusters based on similarity.

• AgglomerativeClustering : Clusters by merging similar points until a single large cluster is formed.

• DBSCAN : Groups points close to each other and filters out noise. No need to specify the number of clusters!

• MeanShift : Shifts data points toward areas of high density to find clusters.
  
  

🌲 6. Ensemble Methods (ensemble)

• RandomForestClassifier : Combines multiple decision trees to make stronger decisions.

• RandomForestRegressor : Predicts continuous values using an ensemble of decision trees.

• GradientBoostingClassifier : Builds trees sequentially, each correcting errors made by the last.

• VotingClassifier : Combines the results of multiple models to make a final prediction.
  
  

📐 7. Metrics (metrics)

Measure your model’s performance:

• accuracy_score : Measures how many predictions your model got right.

• f1_score : Balances precision and recall into a single score.

• roc_curve : Shows the trade-off between true positives and false positives.
  
  

⚙️ 8. Model Selection (model_selection)

• train_test_split : Splits your data into training and test sets.

• KFold : Trains the model in 'k' iterations for better validation.
  
  

🔍 9. Preprocessing (preprocessing)

• StandardScaler : Standardizes your data so it has a mean of 0 and a standard deviation of 1.

• LabelEncoder : Converts text labels into numerical labels (e.g., "cat", "dog").
  
  

🧩 10. Dimensionality Reduction (decomposition)

Dimensionality Reduction helps in simplifying data while retaining most of its valuable information. By reducing the number of features (dimensions) in a dataset, it makes data easier to visualize and speeds up machine learning algorithms.

• PCA (Principal Component Analysis) : PCA reduces the number of dimensions by finding new uncorrelated variables called principal components. It projects your data onto a lower-dimensional space while retaining as much variance as possible.
  
  
  • How It Works: PCA finds the axes (principal components) that maximize the variance in your data. The first principal component captures the most variance, and each subsequent component captures progressively less.
  • Use Case: Use PCA when you have many features, and you want to simplify your dataset for better visualization or faster computation. It is particularly useful when features are highly correlated.

---

🎯 Why Use This Library?

• Learning-First Approach: If you're a beginner and want to understand machine learning, this is the library for you. No hidden complexity, just code.
• No Hidden Magic: Everything is written from scratch, so you can see exactly how each algorithm works.
• Lightweight: Uses only NumPy, making it fast and easy to run.
  
  

🚀 Getting Started

# Clone the repository
git clone https://github.com/adityajn105/MLfromScratch.git

# Navigate to the project directory
cd MLfromScratch

# Install the required dependencies
pip install -r requirements.txt

👨‍💻 Author

This project is maintained by Aditya Jain

🧑‍💻 Contributors

Constributor: Subrahmanya Gaonkar

We welcome contributions from everyone, especially beginners! If you're new to open-source, don’t worry—feel free to ask questions, open issues, or submit a pull request.

🤝 How to Contribute

1. Fork the repository.
2. Create a new branch (git checkout -b feature-branch).
3. Make your changes and commit (git commit -m "Added new feature").
4. Push the changes (git push origin feature-branch).
5. Submit a pull request and explain your changes.
   
   

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.