
AI_Bots
AI_Bots is a comprehensive Applied AI project (COMP-6721) developed by Team AI_Bots. This project consists of two main phases: Phase 1 focuses on a detailed comparative analysis of classical machine learning models (SVM, Decision Trees, and Random Forests) on benchmark datasets, including data preprocessing, hyperparameter tuning, and comprehensive evaluation across multiple metrics. Phase 2 involves the development and training of a custom Convolutional Neural Network (CNN) on the MNIST dataset, focusing on model design, training pipelines, and performance evaluation with visualizations including accuracy and loss curves. The project demonstrates expertise in both classical machine learning and deep learning approaches, with detailed Jupyter notebooks, comprehensive reports, and well-documented code.
Technologies
Features
- Phase 1 - Comparative Analysis: Detailed evaluation of SVM, Decision Trees, and Random Forests on benchmark datasets with comprehensive metrics including accuracy, precision, and recall.
- Data Preprocessing: Advanced data preprocessing techniques including feature engineering, normalization, and handling missing values for optimal model performance.
- Hyperparameter Tuning: Systematic hyperparameter optimization to achieve best model performance across different algorithms.
- Phase 2 - CNN Image Classification: Development and training of a custom Convolutional Neural Network on the MNIST dataset for handwritten digit recognition.
- Model Design: Custom CNN architecture design with proper layer configurations, activation functions, and optimization strategies.
- Performance Visualization: Comprehensive visualizations including accuracy/loss curves, confusion matrices, and performance metrics for model evaluation.
- Jupyter Notebooks: Well-documented Jupyter notebooks with detailed explanations, code comments, and step-by-step analysis for both phases.
- Team Collaboration: Collaborative project with clear task distribution and contributions from team members Shurrab Mohammed, Oleksandr Yasinovskyy, and Huzaifa Mohammed.
- Comprehensive Reports: Detailed project reports (PDF) documenting methodology, results, analysis, and future work ideas for both phases.
- Future Work Extensions: Proposed extensions including XGBoost integration, SVM with RBF kernels, deeper CNN architectures, and data augmentation techniques.
Gallery
