This repository is a collection of diverse machine learning and artificial intelligence projects designed to showcase both fundamental algorithms and practical real-world applications.
Each subdirectory is a self-contained project with its own focus, implementation, and documentation.
A collection of classic and modern machine learning and data mining algorithms implemented for experimentation and learning.
Includes supervised, unsupervised, and deep learning techniques.
Key Implementations:
- A-Priori (Market Basket Analysis)
- Anomaly Detection (Local Outlier Factor, Autoencoders)
- Artificial Neural Networks (feedforward, dropout, early stopping)
- AutoEncoder (dimensionality reduction, anomaly detection)
- Bias-Variance Tradeoff, OLS/WLS regression
- Classification with Imbalanced Datasets
- Decision Trees & Hierarchical Clustering
- K-NN with Cross Validation
- Kernel K-Means Clustering
- LDA (Latent Dirichlet Allocation for topic modeling)
- Linear Regression with L1/L2 regularization
- Singular Value Decomposition (SVD)
- Support Vector Machines (SVM)
Highlights:
- Covers both traditional ML and deep learning
- Focus on clarity and educational value
- Includes visualizations like dendrograms, heatmaps, and confusion matrices
A real-time facial recognition system built with deep learning models.
Combines OpenCV for video processing and DeepFace / FaceNet for face verification.
Features:
- Real-time facial recognition using webcam/video feed
- Cosine similarity for identity verification
- Threshold tuning for robust performance
- Multithreading for efficient video frame processing
- Extensible design for secure identification applications
Tech Stack:
- Python, OpenCV, DeepFace, TensorFlow/Keras
- Multithreading for performance
- Developed and tested in a Linux environment
Applies linear regression to predict Netflix stock closing prices from historical trading data.
Includes both a closed-form regression solution and a custom gradient descent algorithm.
Features:
- Predict Netflix’s stock closing prices using trading volume
- Gradient Descent implementation for iterative optimization
- Graphical visualization of regression performance
- Experimentation with learning rates & epochs
Tech Stack:
- Python, pandas, matplotlib
- Linear Regression & Gradient Descent
Implements a two-layer fully connected neural network from scratch to classify handwritten digits (MNIST dataset).
No pre-built deep learning libraries — all forward/backpropagation and optimization done manually.
Features:
- ReLU activation for hidden layers, Softmax for outputs
- Manual implementation of forward and backward propagation
- Gradient Descent optimization with learning rate decay
- Confusion matrix & per-class accuracy evaluation
- Misclassified digit visualization
Tech Stack:
- Python, NumPy, pandas
- Matplotlib & Seaborn for visualizations
- Scikit-Learn for evaluation metrics
This repository demonstrates a broad spectrum of AI/ML techniques:
- From scratch implementations → Build neural networks and optimization routines manually
- Classic ML algorithms → Regression, clustering, topic modeling, SVMs, etc.
- Practical applications → Stock prediction & real-time facial recognition
It is structured as a learning showcase for:
- Students studying machine learning and AI
- Developers building end-to-end applied ML systems
- Interview and portfolio preparation
Clone the repository:
git clone https://github.com/Philipst77/ML-AI-Projects.git
cd ML-AI-Projects