Analysis of the Price Movement of Cryptocurrencies Using Sentiment Analysis and Machine Learning Methods

Cryptocurrencies complement traditional methods and become new ways of paying for goods and services. Pioneers in accepting cryptocurrency payments include Canada, Malta, Switzerland, Norway, and El Salvador. Investors' interest in cryptocurrencies is growing each year, with the number of registered users on one of the most popular cryptocurrency exchanges, Binance, reaching 7.9 million users in 2019, and exceeding 90 million users by 2022. As a result, the number of companies providing services for trading crypto assets actively increased from 2019 to 2022, with cryptocurrency exchange growth of +219%. By the end of 2022, the number of officially registered crypto exchanges amounted to 525 companies. This study examines the influence of potential investor sentiment on cryptocurrency pricing, specifically posts on Twitter or news portals from 01.01.2019 to 01.01.2023, directly related to cryptocurrencies. Three text data processing methods were used in the research: a model pre-trained on a ready-made dictionary for Bitcoin, Fin-Bert, and Crypto-Bert. By parsing data from the Binance website, a dataset with hourly quotes for 31 studied coins was collected. Over 38,000 posts from Twitter were collected and all processed using the above-listed methods to determine the sentiment of each post. The final step in the study involved identifying the possible influence of online platform sentiment on cryptocurrency profitability using econometric models.