( 09th October 2019 )
Unlike trading the stock market or Forex market, the cryptocurrency market is a more open, worldwide market where discussions are mostly public and easily accessible.
Every day, thousands to millions of conversations focusing on cryptocurrencies take place across multiple platforms. Reddit, Twitter and Telegram are platforms that facilitate these discussions between crypto traders and enthusiasts. Several studies from top universities have shown that there are correlations between the volume and sentiment of discussions on social platforms, and the price action of cryptocurrencies. It is therefore strategic for a trader to follow these conversations and understand the feeling of the crypto market.
But in the face of such an overwhelming amount of data in the form of articles, tweets and blog posts, crypto traders must spend one of their most valuable assets - time - to constantly keep up with market sentiment. This involves spending time reading content, gauging community sentiment and developing personal opinions towards hundreds of cryptocurrencies…
1.1. Psychology and emotions on the Cryptomarket
In trading, psychology is a key concept which is often underestimated. It is critical to detect the feelings surrounding a particular market and get the global sentiment. In cryptocurrencies, market psychology is certainly one of the most important factors to study. Since the discussions about cryptocurrencies are happening online, there is tremendous value in data mining from Twitter, Reddit and other social platforms to detect patterns.
To stay up-to-date with the evolving cryptocurrency market, crypto traders need to monitor and analyze data in real time, so that they can detect trends and hopefully anticipate market shifts. The analysis of web and social media conversations brings a lot of value to better understanding the perception of a cryptocurrency, the specific expectations of the public and the emerging changes in market trends.
But simply analyzing conversations of every Internet user is not enough: you have to track cryptocurrency influencers, specialized media, researchers, journalists and experts to detect emerging trends.
This type of analysis also involves solving several points: knowing where these conversations occur, detecting the insights and trends within these conversations, and analyzing the output data. This output data is so large and complex that automated AI/ML models will be best suited to find patterns and trends. Predicoin solves this problem by tracking social media to provide insights on the crypto market and detecting market sentiment through:
The latest content from crypto news sites.
Twitter and Reddit posts from thousands of crypto influencers
Macro and micro economics/fundamentals of a coin (team, developers, etc.)
Technical indicators on a coin’s price
Popularity and trending characteristics of a coin
Crypto traders need as much information as possible about the crypto market. By tracking and understanding the crypto market sentiment and crypto traders’ emotions, traders can utilize another layer of information when developing their trading strategy.
Since sentiment analysis of online content boils down to a big data analytics problem, Predicoin to crypto traders derive insight from the market, and save a lot of time doing so. Crypto traders will get insight to the macro and micro trends from each cryptocurrency and be able to stay up-to-date on the market sentiment.
There have been previous attempts to utilize sentiment from tweets to predict fluctuations in the price of bitcoin. Coliannni et al. reported 90% accuracy in predicting price fluctuations using similar supervised learning algorithms; however, their data was labeled using an online text sentiment API. Therefore, their accuracy measurement corresponded to how well their model matched the online text sentiment API, not the accuracy in terms of predicting price fluctuations. Similarly, Stenqvist and Lonno utilized deep learning algorithms, on a much higher frequency time scale of every 30 minutes, to achieve 79% accuracy in predicting bitcoin price fluctuations using 2.27 million tweets. Neither of these methods use data labeled directly based on price fluctuations, nor did they analyze the average size of price percent increases and percent decreases their models were predicting.
More classical approaches of using historical price data of cryptocurrencies to make predictions have also been tried. Hegazy and Mumford achieved 57% accuracy in predicting the actual price using supervised learning methods. Jiang and Liang utilized deep reinforcement learning to manage a bitcoin portfolio that made predictions on price. They achieved a 10x gain in portfolio value. Last, Shah and Zhang utilized Bayesian regression to double their investment over a 60-day period. None of these methods utilized news or social media data to capture trends not apparent in the price history data.
Comentarios