Machine Learning Models of Crypto Currency Trends Using Attention-based Sentiment Analysis

Abstract

The field of machine learning and the burgeoning world of cryptocurrency share many commonalities, and their integration presents opportunities for valuable insights. Machine learning utilizes algorithms to analyze and predict patterns based on vast datasets. In contrast, cryptocurrency is a digital or virtual currency that leverages encryption techniques for fund transfers and unit generation regulation. By examining the intersection of these two fields, researchers can gain valuable insights into cryptocurrency trends and market behavior. One approach to exploring this intersection involves using attentionbased sentiment analysis in machine learning models. Attention-based models, a type of neural network, identify the most relevant aspects of an input to the output. Sentiment analysis determines and categorises text data’s emotional tone, such as news articles or social media posts. Researchers can use these techniques to develop machine learning models that analyze cryptocurrency-related data sources, such as social media or news articles, to identify trends and sentiments. This type of analysis has practical applications for investors, traders, and policymakers interested in cryptocurrency markets. Sentiment analysis can be used to identify bullish or bearish sentiment among market participants, potentially providing valuable insights into market trends. In contemporary times, machine learning models have emerged as a compelling tool for anticipating future market trends. This is particularly valuable for informed decision-making concerning investment and policy measures. However, using machine learning models to analyze cryptocurrency data presents several noteworthy challenges. The foremost obstacle is the high degree of volatility characterizing the cryptocurrency market. This dynamic makes it difficult to devise precise predictions with machine learning models. Additionally, the decentralized nature of cryptocurrencies and the lack of data standardization pose considerable challenges to data collection and analysis processes. Despite these challenges, the integration of machine learning and cryptocurrency has the potential to yield valuable insights into market behavior and trends. Attention-based sentiment analysis presents a promising approach to exploring this intersection. Further research in this area can contribute to understanding cryptocurrency markets and their broader financial system implications.

Date of Award	22 Jul 2024
Original language	American English
Supervisor	Ibrahim Elfadel (Supervisor)