This repository contains a collection of three independent machine learning projects developed by me. Each project demonstrates different ML techniques and applications.
A Convolutional Neural Network (CNN) approach to both classify and locate emojis within images.
A Natural Language Processing (NLP) project using Naive Bayes classification to determine sentiment in tweets.
A recommendation system using collaborative filtering techniques to suggest books based on user preferences.
This project combines image classification with object localization to identify emoji types and their positions within images. The system works under the assumption that each image contains exactly one emoji.
- Model Architecture: Custom CNN with both classification and regression outputs
- Classification Task: Identifying which emoji appears in the image
- Localization Task: Predicting bounding box coordinates (x, y, width, height)
- Implementation: Implemented in Python using TensorFlow/Keras
- Single-stage detection and classification pipeline
- Simplified approach to object detection without requiring complex frameworks like YOLO or R-CNN
- Demonstrated balance between computational efficiency and accuracy
Refer to the emoji_localization_and_classification.ipynb notebook for implementation details, model training, and evaluation.
A text classification project that automatically categorizes tweets based on their sentiment (positive, negative, or neutral). This solution helps businesses understand customer opinions expressed on social media without manual review.
- Algorithm: Naive Bayes classifier
- Text Processing: Tokenization, stop word removal, stemming/lemmatization
- Feature Extraction: Bag-of-words and TF-IDF representations
- Implementation: Python with scikit-learn and NLTK
- Brand monitoring and reputation management
- Customer feedback analysis
- Market research and competitive analysis
- Real-time sentiment tracking during events or product launches
See the twitter_sentiment_analysis.ipynb notebook for the complete implementation, including data preprocessing, model training, and evaluation metrics.
A recommendation engine that suggests books to users based on collaborative filtering techniques. The system analyzes patterns in user preferences and book ratings to make personalized recommendations.
- Algorithm: Collaborative filtering
- Similarity Metrics: Cosine similarity between user vectors
- Implementation: Python with pandas, numpy, and scikit-learn
- User-based collaborative filtering to find similar users
- Item-based collaborative filtering to find similar books
- Handling of the cold-start problem for new users
- Evaluation using standard recommendation system metrics
This system provides significant utility to e-book platforms and online libraries by:
- Increasing user engagement through personalized recommendations
- Improving discovery of less popular but relevant titles
- Enhancing user experience and satisfaction
Explore the book_recommender_system.ipynb notebook for implementation details, including data preparation, model development, and recommendation generation examples.
machine-learning-projects/
├── emoji_localization_and_classification.ipynb
├── twitter_sentiment_analysis.ipynb
├── book_recommender_system.ipynb
└── README.md
- Python
- TensorFlow
- scikit-learn
- NLTK
- pandas
- numpy
- matplotlib
- Jupyter Notebook / Google Colab
Detailed requirements for each project are specified in their respective notebooks.
Potential enhancements for each project:
- Emoji Project: Extend to multi-emoji detection, add support for custom emojis
- Sentiment Analysis: Implement more advanced models (BERT, Transformers), add multi-language support
- Book Recommender: Incorporate content-based features, develop hybrid recommendation approach