Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
102 tokens/sec
GPT-4o
59 tokens/sec
Gemini 2.5 Pro Pro
43 tokens/sec
o3 Pro
6 tokens/sec
GPT-4.1 Pro
50 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Anticipating cryptocurrency prices using machine learning (1805.08550v4)

Published 22 May 2018 in physics.soc-ph, cs.LG, q-fin.GN, q-fin.ST, and q-fin.TR

Abstract: Machine learning and AI-assisted trading have attracted growing interest for the past few years. Here, we use this approach to test the hypothesis that the inefficiency of the cryptocurrency market can be exploited to generate abnormal profits. We analyse daily data for $1,681$ cryptocurrencies for the period between Nov. 2015 and Apr. 2018. We show that simple trading strategies assisted by state-of-the-art machine learning algorithms outperform standard benchmarks. Our results show that nontrivial, but ultimately simple, algorithmic mechanisms can help anticipate the short-term evolution of the cryptocurrency market.

Anticipating Cryptocurrency Prices Using Machine Learning

The paper "Anticipating Cryptocurrency Prices Using Machine Learning” investigates the efficacy of leveraging machine learning algorithms to predict cryptocurrency prices, a challenging task given the market’s inherent volatility and inefficiencies. The authors focus on exploiting these inefficiencies to generate abnormal profits, analyzing data from 1,681 cryptocurrencies between November 2015 and April 2018.

The paper explores three distinct models for price prediction: two based on gradient boosting decision trees and one on long short-term memory (LSTM) recurrent neural networks. A baseline method using a simple moving average strategy is implemented for comparison. The paper posits that machine learning methodologies, particularly those incorporating time-series forecasting techniques like LSTM, can outperform traditional financial analysis methods in predicting short-term price fluctuations.

Key Findings

  1. Model Performance: The LSTM model demonstrated superior results in terms of return on investment across all considered periods, outperforming both the gradient boosting decision trees models and the baseline method. The ability of LSTMs to capture long-term dependencies in sequence data gives them a distinct advantage in the field of cryptocurrency markets.
  2. Short-term Strategies: The methods based on gradient boosting decision trees were optimized for short-term prediction windows, indicating their suitability for exploiting immediate market inefficiencies. These models offered enhanced interpretability of the features driving market predictions, which are crucial for practical trading strategies.
  3. Cumulative Return: In theoretical scenarios where the available supply is not limited and trades do not influence market prices, the cumulative return performance of these models was substantial, particularly the LSTM-based approach. However, the authors also consider realistic scenarios with transaction fee constraints and liquidity limitations, where returns were still favorable, albeit reduced.
  4. Feature Importance: Price and ROI history were identified as significant predictors within the random forest models, underscoring the impact of recent price movements on short-term forecasts.
  5. Impact of Market Growth: Price predictions expressed in Bitcoin rather than USD provided more accurate results by circumventing challenges posed by the market’s overall growth trend during the paper period.

Implications and Future Directions

The paper contributes to the ongoing exploration of artificial intelligence in financial markets, highlighting practical applications for machine learning in cryptocurrency trading. The findings suggest that more sophisticated neural network architectures capable of learning complex patterns over longer periods could lead to further improvements in forecast accuracy and trading strategies. Additionally, the examination of market inefficiencies invites broader discussions on portfolio management and risk assessment in volatile markets.

The authors identify several potential avenues for further research, such as integrating social media sentiment analysis with predictive models to capture external factors influencing cryptocurrency prices. Moreover, enhancing the models to incorporate real-time intra-day trading data could refine predictions and offer traders additional tools for managing investments.

In conclusion, the research underscores the transformative potential of machine learning in cryptocurrency markets, providing a foundation for subsequent studies and practical applications to optimize trading strategies amidst the complexities of digital currency fluctuations.

User Edit Pencil Streamline Icon: https://streamlinehq.com
Authors (4)
  1. Laura Alessandretti (26 papers)
  2. Abeer ElBahrawy (6 papers)
  3. Luca Maria Aiello (60 papers)
  4. Andrea Baronchelli (82 papers)
Citations (167)