Crypto Trader Classifier with Logistic Regression
December 13th, 2024

Introduction

As the cryptocurrency market grows increasingly dynamic, identifying profitable traders has become essential for enhancing investment strategies. This research project introduces a logistic regression model trained to classify crypto wallets into three categories: Good Trader, Average Trader, and Bad Trader. The model utilizes key features derived from trading activity data to achieve a robust classification. Features Used The following features were extracted and used to categorize traders effectively:

  • Base Cumulative Return: The total return generated by the trader over time.

  • Portfolio Return: The percentage return on the trader’s entire portfolio.

  • Daily Sharpe Ratio: A measure of risk-adjusted returns.

  • Number of Trades: The total trades executed by the wallet.

  • Unique Tokens Traded: The diversity of assets traded by the wallet.

Features from the Flipside Query
Features from the Flipside Query

Methodology

Data Collection The data was collected using Flipside Crypto’s SQL terminal. Initial SQL queries targeted traders with a trading volume exceeding $10 million. These traders were tracked for their trading activities, and the resulting data was exported as CSV files. Python was then used for:

  • Statistical analysis

  • Feature engineering

  • Training the logistic regression model

    Correlation Analysis

    Correlation analysis was conducted to identify the most relevant features for prediction. Correlation values provide insights into how closely variables are related:

  • +1: Strong positive correlation

  • -1: Strong negative correlation

  • 0: No correlation

Correlation Coefficient
Correlation Coefficient

In this study, the strongest correlations were observed for:

  • Portfolio Return: Correlation coefficient of 0.2718.

  • Base Cumulative Return: Correlation coefficient of 0.2387. These features were prioritized in the modeling process.

Logistic Regression Model Overview Logistic regression is a classification algorithm used to predict categorical outcomes based on input features. In this study, the model’s target variable was multi-class (“Good Trader,” “Average Trader,” “Bad Trader”), allowing it to classify traders effectively.

Model Training Using the extracted features, the model learned the relationship between these variables and the trader categories. The training process involved optimizing the model’s parameters to maximize predictive accuracy.

Performance Metrics The logistic regression model achieved an accuracy score of 97.87%. This indicates a 97% probability that the classification of a trader is correct, showcasing the model’s high reliability and effectiveness.

Accuracy Score
Accuracy Score

Results and Insights

  • Correlation Analysis: Highlighted the most significant features influencing trader classification.

  • Model Accuracy: Demonstrated the model’s robustness in categorizing traders into meaningful groups.

  • Feature Importance: Emphasized the role of portfolio return and base cumulative return in predicting trader performance. Conclusion This project demonstrates the power of machine learning in analyzing on-chain trading activities and classifying crypto traders. By leveraging logistic regression and key trading metrics, the model provides actionable insights for identifying high-performing traders. Applications

  1. Copy-Trading: Replicating strategies of "Good Traders."

  2. Risk Management: Mitigating exposure to "Bad Traders."

  3. Portfolio Optimization: Enhancing returns through data-driven decisions. References

Subscribe to apostleoffinance
Receive the latest updates directly to your inbox.
Mint this entry as an NFT to add it to your collection.
Verification
This entry has been permanently stored onchain and signed by its creator.
More from apostleoffinance

Skeleton

Skeleton

Skeleton