Machine Learning Practice Repository

This repository contains my machine learning practice notebooks, datasets, and small projects. The focus is on learning core concepts by implementing models, exploring real datasets, and gradually moving from basic algorithms to applied machine learning workflows.

Repository Contents

Regression

• linear-regression.ipynb
  Linear regression from scratch for a single feature.

• linear_regressiovn_multivariate.ipynb
  Multivariable linear regression with feature scaling and gradient descent.

• medical_insurance.ipynb
  Regression workflow for predicting medical insurance charges.

• insurance.csv
  Dataset used in the medical insurance notebook.

Classification

• logistic_regression.ipynb
  Logistic regression from scratch for binary classification.

• breast_cancer.ipynb
  Classification workflow using the Breast Cancer Wisconsin dataset.

• data.csv
  Dataset used in the breast cancer notebook.

• decision-tree-on-heart-disease.ipynb
  Decision tree experiments on a heart disease dataset.

• decision-tree-from-scratch.ipynb
  Basic decision tree implementation without relying on scikit-learn's tree model.

• k-nearest-neighbours.ipynb
  KNN classification on the Breast Cancer Wisconsin dataset with accuracy analysis across different K values.

• naive-bayes.ipynb
  Gaussian Naive Bayes classifier on the Breast Cancer Wisconsin dataset with evaluation and ROC curve.

Neural Networks

• neural_net_with_mnist_from_scratch.ipynb
  Basic neural network implementation for MNIST-style digit classification.

• digit-recognizer/
  Dataset files for digit recognition work, including train.csv, test.csv, and sample_submission.csv.

Clustering

• k-means.ipynb
  K-Means clustering practice using customer segmentation style data.

• k-means-from-scratch.ipynb
  K-Means clustering implemented from scratch.

• Mall_Customers.csv
  Dataset used for customer clustering practice.

Anomaly Detection

• anamoly-detection.ipynb
  Network anomaly detection practice using security-related network data.

• embedded_system_network_security_dataset.csv
  Dataset used for anomaly detection experiments.

• Train.txt and Test.txt
  Additional network anomaly detection training and testing data.

Recommender Systems

• recommender_system.ipynb
  Anime recommendation system practice using popularity-based recommendations, content-based filtering, and collaborative filtering.

• anime.csv
  Anime metadata, including title, genre, type, episodes, rating, and members.

• rating.csv
  User-anime rating data used for recommender system experiments.

Other Files And Folders

• archive/
  Older or extra project material kept outside the main notebook flow.

• build-your-claude-code-from-scratch/
  Separate coding project folder kept in the workspace.

• .gitignore and .gitattributes
  Repository configuration files.

Topics Covered

• Linear regression
• Multivariable regression
• Logistic regression
• Decision trees
• Neural networks
• Gradient descent
• Feature scaling
• Classification workflows
• Regression workflows
• K-Means clustering
• Unsupervised learning
• Anomaly detection
• K-Nearest Neighbors (KNN)
• Naive Bayes
• Recommender systems
• Working with real datasets

Current Learning Path

The repository currently follows this progression:

1. Linear regression from scratch
2. Multivariable regression
3. Logistic regression
4. Real-world classification and regression notebooks
5. Decision trees
6. Neural networks and digit recognition
7. Unsupervised learning
8. K-Means clustering
9. Anomaly detection
10. K-Nearest Neighbors (KNN)
11. Naive Bayes
12. Recommender systems

Recommender System Direction

The recommender system project uses the anime dataset and is organized into three recommendation approaches:

1. Popularity-based recommender
   Recommends anime using overall rating and member count.

2. Content-based recommender
   Recommends anime similar to a selected title using item features such as genre and type.

3. Collaborative filtering recommender
   Recommends anime using user rating behavior and cosine similarity between anime.

The notebook also includes data loading, missing value handling, exploratory data analysis, and user-level recommendation examples.

Note

This is an active learning repository. The notebooks, datasets, and project structure may change as the work becomes more complete and better organized. Next step is SVM and whatever was not covered in the ML spec and then after that RAG, Agents, MCP, etc