Kinh tế, tài chính - Ngân hàng và kế toán, kiểm toán trong bối cảnh chuyển đổi số 2022 - Hội thảo khoa học: Phần 2

Nối tiếp nội dung phần 1, phần 2 cuốn "Hội thảo khoa học: Kinh tế, tài chính - Ngân hàng và kế toán, kiểm toán trong bối cảnh chuyển đổi số 2022" tiếp tục trình bày các bài viết về: ảnh hưởng của các yếu tố văn hóa, truyền thống đến đạo đức nghề nghiệp kế toán - kiểm toán trong doanh nghiệp; phân tích dữ liệu lớn: cơ hội và thách thức đối với nghề kế toán; một số giải pháp thu hẹp khoảng cách kỳ vọng trong kiểm toán;... Mời các bạn cùng tham khảo!
Nội dung trích xuất từ tài liệu:
Kinh tế, tài chính - Ngân hàng và kế toán, kiểm toán trong bối cảnh chuyển đổi số 2022 - Hội thảo khoa học: Phần 2 393 BITCOIN PRICE MOVEMENT PREDICTION BY NEWS SENTIMENT USING MACHINE LEARNING APPROACH Phan Huy Tam 1 First and Corresponding Author, Ph.D. Candidate, Faculty of Finance and Banking, University of Economics and Law, VietNam National University, Ho Chi Minh City, Vietnam. Email: tamph@uel.edu.vn Abstract Bitcoin is the most popular cryptocurrency nowadays, the community accepting this kind of currency as a mean of payment has been developed rapidly. However, Bitcoin historical price showed a significant fluctuation in both short-term and long-term valuations. Besides, with the evolve of technology, an enormous amount of data in various forms is now available for researchers to mine. Google News is increasing used as a tool to search and gather information and making investment decision regarding Bitcoin for investors. As a result, quickly access and understand the hidden knowledge in news data could provide advantages for traders. This paper presents an approach in predicting Bitcoin price movement by using text-based data observed from Google News search platform utilized by sentiment analysis and machine learning algorithms. The test results using sentiment analysis and different machine learning algorithms shows that Decision Tree and Random Forest have reasonably good balanced prediction, while Logistic Regression and Support Vector Machine have high AUC score, but the prediction is unbalanced between classes, Naïve Bayes and KNN experienced ineffective results. This research contributes both theoretically and practically, helping investors to have more investment decision making tool and better understand the impact of news on cryptocurrency prices. Keywords: Bitcoin Movement, Machine Learning, Google News, Sentiment Analysis 1. Introduction Since the launch of Bitcoin, this type of cryptocurrency has become a superstar and gain trust around an increasing community. Bitcoin has the highest market capitalization in term of cryptocurrency and considered to be called as “digital gold”. This achievement is based on the faith of investors on the future of Bitcoin and the underneath technology (Prajapati, 2020). In recent days, the huge demand for Bitcoin trading states the interest of the market on cryptocurrencies and the blockchain technology in general. Bitcoin is a decentralized cryptocurrency which does not required any central authority such as bank and could be transferred and used in many purposes through a peer-to-peer network. The circulation of Bitcoin uses a special algorithm which was introduced under a name of Satoshi Nakamoto, all transactions are public on a system of distributed ledge to ensure the transparency and anonymity at the same time (Nakamoto, 2008). Figure 1 shows the historical price of Bitcoin since its inception. The graph states a sharp increase in Bitcoin value since the first released in 2009 to the peak around $20,000 in 2017. This indicates that Bitcoin is a good term of investment and attracted a vast amount of capital around the world. The number of Bitcoin user was around 4 million when it reach the highest value of all time in 2017 (Hileman & Rauchs, 2017). Even if the Bitcoin owner used it as a mean of payment or treat it as an investment, the fluctuation in value of Bitcoin is uncertainty. This new store of value is considered as an unique @ Trường Đại học Đà Lạt 394 asset and behaves in ways similar to both standard financial asset and a speculative one (Kristoufek, 2015). This makes Bitcoin prices extremely hard to predict (Abraham, Higdon, Nelson, & Ibarra, 2018). Since people’s trust is involved in the rise of the cryptocurrency market, the sentiment of the general population does make a huge impact on the future of the cryptocurrency market capitalization (Kristoufek, 2015). The application of text-based data with sentiment analysis to predict future trend of cryptocurrencies is widely used across academic field. The source of data varies from journal news headlines, content, or social media posts like Facebook, Twitter or even anonymous community as Reddit (Abraham et al., 2018; Gerritsen, Lugtigheid, & Walther, 2022; Huang et al.; Kristoufek, 2015; Mittal, Dhiman, Singh, & Prakash, 2019; Vo, Nguyen, & Ock, 2019). In that, Google News is a very powerful search engine to collect news posted from various sources around the world. Furthermore, this nice source of text-based data is capable to crawl news based on selected keywords as well as the other information related to the text-based data such as posted date, author… In this research, the author uses text-based data to feed in the sentiment analysis and then combine with historical price and volume of Bitcoin. The processed data is used to train machine learning model to investigate the correlation between these variables. This research aims to examinate the prediction power of newspaper headlines which are collected from Google News platform, using the sentiment analysis and various machine learning algorithms. This type of non-quantifiable data showed a potential effect in predicting cryptocurrencies trend direction by capturing the overall market point of view about the volatility of the cryptocurrencies. The test results provide evidence of the relationship between text-based data from news headlines to short-term movement of Bitcoin price in both academic and pra ...