CryptoClustering Challenge

DataClass Module 19 – Unsupervised Learning
EdX/UT Data Analytics and Visualization Bootcamp
Cohort UTA-VIRT-DATA-PT-11-2024-U-LOLC
Author: Neel Agarwal

---

Table of Contents

1. Project Overview
2. Installation & Setup
3. Unsupervised Learning Methodology
4. Directory Structure
5. Rubric & Evaluation
6. Usage Notes & Limitations
7. Credits & References

---

Project Overview

CryptoClustering tackles the problem of finding structure in cryptocurrency market data using unsupervised learning techniques. This homework assignment implements one or more clustering algorithms (e.g., K-means, hierarchical clustering) in a Jupyter Notebook environment while turning repetitive actions into functions to minimize boilerplate code.

Key objectives:

• Preprocess crypto-market data to normalize and prepare for clustering.
• Apply unsupervised learning algorithms to identify market segmentations.
• Evaluate clustering results based on the provided rubric criteria.
• Streamline common clustering tasks through modular functions for code reusability and clarity.

---

Installation & Setup

1. Clone the Repository

   git clone https://github.com/yourusername/CryptoClustering.git
   cd CryptoClustering

2. Set Up Virtual Environment (Recommended)

   python -m venv venv
   # On Unix/MacOS:
   source venv/bin/activate
   # On Windows:
   venv\Scripts\activate

3. Install Required Dependencies

   pip install -r requirements.txt

4. Run the Jupyter Notebook

   Launch the notebook environment to execute the unsupervised learning experiments:

   jupyter notebook Crypto_Clustering.ipynb

Note

Ensure that you have installed all required Python packages (such as pandas, numpy, scikit-learn, and matplotlib) as listed in your requirements.txt.

---

Unsupervised Learning Methodology

• Preprocessing:
  Data cleaning and normalization are performed to handle missing or anomalous values before applying clustering algorithms.

• Clustering Algorithms:
  The notebook includes implementations of:

  • K-means Clustering: To partition data into optimal groups.
  • Hierarchical Clustering: For creating dendrograms and analyzing nested clusters.

• Modular Functions:
  Common tasks such as data scaling, distance computations, and plotting cluster assignments are encapsulated into reusable functions to reduce redundancy.

• Evaluation:
  Clustering quality is assessed using metrics (e.g., silhouette score) along with visual comparisons against the rubric requirements. Detailed evaluation criteria are outlined in the attached rubric.

---

Directory Structure

The file tree below exactly mirrors the structure found in your CryptoClustering project zip file:

CryptoClustering/
├── .gitignore
├── Crypto_Clustering.ipynb
├── README.md
├── Resources/
│   └── crypto_market_data.csv
├── __init__.py
└── lib.py

---

Rubric & Evaluation

The project evaluation was guided by the rubric provided in the HTML rubric. Key criteria include:

• Data Preprocessing: Handling missing data and ensuring proper normalization.
• Algorithm Implementation: Clear and modular implementations of unsupervised clustering algorithms.
• Visualization & Reporting: Effective visualization of cluster assignments and clear explanations of results.
• Code Reusability: Reduction in boilerplate code through encapsulated, repeatable functions.
• Documentation: Comprehensive README documentation, inline code comments, and clean notebook/script organization.

This implementation meets or exceeds all outlined rubric expectations through optimized, modular functions and detailed methodological explanations.

---

Usage Notes & Limitations

• Usage Notes:

  • Run the main notebook (Crypto_Clustering.ipynb) to see interactive clustering analyses.
  • Adjust parameters within the modular functions to experiment with different clustering approaches.
  • Ensure that the crypto_market_data.csv file is located in the Resources folder to avoid errors.

• Limitations:

  • The current dataset is limited to the provided CSV file. Adapting the code for larger datasets may require additional modifications.
  • Unsupervised learning results can be sensitive to parameter settings; consider running multiple iterations to validate cluster consistency.

---

Credits & References

• pandas Documentation
• scikit-learn Documentation
• Matplotlib Documentation
• Unsupervised learning theories and best practices from core machine learning literature.
• Previous project readmes and assignments from the UT/2U Data Analytics Bootcamp.