This repository serves as a showcase of machine learning skills, featuring two distinct projects. The first project focuses on image processing and recognition, utilizing the MNIST database to train and evaluate a random forest model capable of recognizing handwritten digits. The second project employs various unsupervised learning methods to analyze a dataset containing arrest statistics per 100,000 residents for assault, murder, and rape in each of the 50 US states in 1973.
The code for data preprocessing, model training, tuning, and evaluation is found in /projects_code.
Develop and evaluate a random forest model for recognizing handwritten digits using the MNIST database.
Apply unsupervised learning techniques to analyze arrest statistics data for assault, murder, and rape in US states in 1973.
-
Clone the repository:
git clone https://github.com/AlexUOM/Machine-learning-projects.git cd Machine-learning-projects -
Navigate to the project directory:
cd projects_code -
Execute the main script for model training and evaluation:
python project1_image_recognition.py
or
python project2_arrests_analysis.py
-
Python 3.x
-
Required Python libraries:
numpy==1.21.0 pandas==1.3.0 tqdm==4.61.1 matplotlib==3.4.3 scikit-learn==0.24.2 seaborn==0.11.2 keras==2.4.3
-
MNIST Digit Recognition:
- Pros: Achieves high accuracy, suitable for digit recognition tasks.
- Cons: May be computationally intensive during training.
-
US States Arrest Statistics Analysis:
- Pros: Reveals patterns and relationships in the dataset.
- Cons: Interpretability may vary depending on the clustering method used.
Contributions, feedback, and suggestions are welcome. Please refer to the /CONTRIBUTING.md file for guidelines.
Thank you to the developers of the MNIST database and contributors to the US states arrest statistics dataset. Special thanks to the machine learning community for inspiration and support.