3 Applications of AI & Machine Learning in Cryptocurrency Trading and Investing

[Davidvictorson]

I am a big fan of Artificial Intelligence (AI). Especially machine learning (ML). According to researchers , "Artificial intelligence (AI) refers to the simulation of the human mind in computer systems that are programmed to think like humans and mimic their actions such as learning and problem-solving". ML is one of the branches of AI.

I get stoked when I create, read or watch interesting AI projects.  For example, Google’s self-driving car project, and Tesla’s “autopilot” feature are powered by AI technology. Also,  Google's,  Bing's, Yahoo's, and DuckDuckGo's internet search engines make us of AI. YouTube, Netflix, Amazon, and eBay utilize AI for recommending videos and movies to watch, and products to buy. In addition, Alexa and Siri are AI-programmed, thanks to advances in natural language processing (NLP). AI in cryptocurrency and blockchain is not left out.

Cryptocurrency and Blockchain have been increasingly gaining traction with Bitcoin and Ethereum prices reaching an all-time high in 2021. These gains are not unrelated to the application of AI in cryptocurrency trading and investing. This article aims to briefly review four(4) applications of AI in cryptocurrency trading and investing. I will also introduce the research efforts that are using AI techniques in crypto-related trading and investing.

1. Sentiment Analysis
There’s a growing number of retail and institutional investors trading in cryptocurrencies such as Bitcoin and Ethereum. As of the writing of this article, there are 305 cryptocurrency exchanges tracked on CoinMarketCap .

In cryptocurrency trading, AI techniques are involved in building systems that help investors make accurate decisions about whether to invest in a crypto asset or not.

Typing “machine learning sentiment analysis and natural language processing in cryptocurrency trading and investment” on Google search engine yields over 1.5 million search results, but let’s draw our attention to academic efforts.

In their research, Kumar analyzed crypto news sentiment to predict bitcoin prices. Also, Vo, Nguyen, and Ock, research paper  analyzed the ability of news data to predict the price fluctuations of Ethereum in terms of market capitalization.

Sentiment analysis is the process of transforming raw texts into bags of words and classifying them into positive, negative, or neutral sentiments. This analysis uses two AI techniques known as Natural Language Processing (NLP) and ML. Sentiment analysis helps gauge how individual opinions affect the market price of an asset or investment. The data for sentiment analysis can be collected from social media sources particularly Twitter, latest industry news through media outlets and blogs.  Using natural language processing algorithms model, the researchers were able to directly predict price direction by indicating whether to buy, sell, or hold.

Takeaway
Using natural language processing algorithms in sentiment analysis is a viable way to identify the public moods for cryptocurrency fluctuations.


2. On-chain Analysis
On-chain analysis involves tracking information on cryptocurrency transactions in a blockchain network. It is a public digital ledger of transactions that occur on the blockchain. Tracking this information is possible due to the publicly available blockchain transactions and market data. The goal of on-chain analysis is to improve trading and investment decisions.

Some of the metrics that are from the on-chain data include; wallets address > 1, >10, >100 coins, transaction count, daily active addresses,  total addresses, total new addresses, hash rate, transactions rate, transfers, gas price, count,  exchange withdrawals, etc.

Several AI techniques can be employed in analyzing on-chain metrics to determine which crypto-assets to make part of one's portfolio or not. For example, researchers developed a self-adapting algorithm in deep learning models to predict the price of Ethereum.

Takeaway
AI approaches to analyzing on-chain metrics is a efficient way to predict the value of crypto assets


3. Deep Reinforcement Learning
Reinforcement learning is a type of ML. It is a framework whereby an agent learns how to interact with the environment through experience, trial and error, and positive and negative feedback. Sattarov et al. authored a research paper  published in February 2020 that applies the deep reinforcement learning (DRL)(neural networks combined with reinforcement learning) model for trading that tries to maximize short-term profits in the cryptocurrency market.

Their model was tested with three—Bitcoin (BTC), Litecoin (LTC), and Ethereum (ETH)—crypto coins’ historical data. The result of the application of DRL on Bitcoin, Litecoin (LTC), and Ethereum (ETH) showed that the investor got 14.4%, 74%, and 41% net profits within one month respectively. The authors proved that through machine learning techniques traders and investors can choose when to buy, hold or sell their crypto assets.

Takeaway
Researchers have created deep reinforcement learning systems that recommend trading options that are useful for increasing the trader's investment.

Thank you for reading. There are even more applications of AI in cryptocurrency trading and price forecasting than I have covered in the article.  I would be delighted to know what you think about  AI and its application  in the world of cryptocurrencies and blockchain.


References
https://en.wikipedia.org/wiki/Artificial_intelligence
https://towardsdatascience.com/machine-learning-in-the-world-of-blockchain-and-cryptocurrency-68651ebaecd7
https://time.com/nextadvisor/investing/cryptocurrency/bitcoin-price-history/
https://time.com/nextadvisor/investing/cryptocurrency/ethereum-price-history/
https://www.nature.com/articles/s41374-020-00514-0
https://www.researchgate.net/publication/357042521_An_On-Chain_Analysis-Based_Approach_to_Predict_Ethereum_Prices
https://coinmarketcap.com/rankings/exchanges/
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3913652
http://www.ijke.org/vol5/116-MK032.pdf

[tranthidung]

AI and Machine Learning can be useful for trading, especially bot trading. Because you can code and do back test on TradingView ie. then apply it in the market.

With investment, it is more related to project's fundamentals so AI or Machine Learning is not necessary. It requires different approach than trading and it's clearly that investors should have different traits than traders.

Generally investors usually have a happier life, less stress than traders and they have more free time to enjoy their own lives.

[hugeblack]

Cryptocurrency market is small and new technology, and therefore artificial intelligence will not achieve the desired success due to several things, the most important of which are:

 - Extremism of values: As I mentioned, the cryptocurrency market has extreme values ​​and therefore it is difficult to find a similar pattern.
 - Small market capacity: We are talking about total investments of less than 2 trillion dollars, which is still a small number compared to other sectors.
 - Speculation on centralized platforms: Most of the trading volumes for many altcoins are not real.
 - Absence of regulation: The market is open to all countries and all options.
 - Lack of data: It's true that it's more than 10 years old, but data is still scarce, making AI an excellent tool for analysis.

Their model was tested with three—Bitcoin (BTC), Litecoin (LTC), and Ethereum (ETH)—crypto coins’ historical data. The result of the application of DRL on Bitcoin, Litecoin (LTC), and Ethereum (ETH) showed that the investor got 14.4%, 74%, and 41% net profits within one month respectively. The authors proved that through machine learning techniques traders and investors can choose when to buy, hold or sell their crypto assets.

So 41% of net profits within one month is general term that does not mean the success of the method, it can be obtained without doing anything.

[zaesvlas]

To be honest, I don’t really believe in such training options. The best learning option is practice. Convinced that this option works.

[Flexystar]

That is deep. Definitely AI and machine learning is future of the world. I am not into technical field of programming or Information technology as whole but I love reading about these topics. One thing that I learnt from reading reddit topics and all, big companies like Google, Apple, IBM all works with AI to enhance their user experience and do the profiling.

No wonder this power will surely be used in the crypto currency learning real soon. I’m not sure about how much costing it would have but the development would be off the roof.

I’m not sure if it is just for trading and investment but surely other development like chain mining, pow in smarter and energy efficient ways.

[crwth]

It's been a while since this topic was posted, and it's not bad for information for AI and ML to be discussed because I think there is potential with that, and ordinary people that manage to take advantage of this could benefit from it for sure.

For the Sentiment Analysis:
If you had large enough data and checked all the different parameters that there could be, you could arrive at some conclusion that it would be wise to invest in the results. Like it's going in with the wave of emotions and "sentiments" of people, and if you could detect or predict the initial start of it, you might have some breakthrough with AI and ML with this.

For On-Chain Analysis:
This would be great if you have a specific chain that you look forward to investing in. Either you will have reinforcement on what you believe in, or it will be something that you would advise yourself to go out. I think that's one of the ways it could help you as an investor.

For Deep Reinforcement Learning:
This requires more data; maybe if this could be combined with the other two analyses above, it would be unstoppable and have a greater chance of getting the correct prediction.