黃啟祐Huang, Chi-Yo吳宜庭Wu, I-Ting2023-12-089999-12-312023-12-082023https://etds.lib.ntnu.edu.tw/thesis/detail/36df28d057c7ccfe1a5a991837645865/http://rportal.lib.ntnu.edu.tw/handle/20.500.12235/120726隨著加密貨幣的問世與日益普及，使用與投資加密貨幣，已成為新的趨勢。但因虛擬貨幣價格的高波動性，投資風險極高，如何精確預測虛擬貨幣價格，非常值得關注。社群媒體的問世與普及，改變了大部份使用者交換資訊的方式。人們除了使用搜尋引擎之外，也可以透過Facebook或Twitter等平台，張貼或搜尋他人的意見與評論。因此，社群媒體上的大量留言，成為大數據分析之標的，對加密貨幣之分析亦然。之前常見社群媒體言論對股票市場之影響分析及預測，但虛擬貨幣為新型態之投資標的，社群媒體言論對比特幣值漲跌幅之衝擊，較少學者著墨。因此本研究擬爬取主要社群媒體之關於加密貨幣之推文，進行文本情感分析，並且同步爬取具代表性之個體經濟與總體經濟指數，將數據一般化（Normalization）之後，使用k-平均群落分析法（k-means cluster analysis）將每一變數分群，再導入優勢約略集合（Dominance Rough Set Approach, DRSA)及形式概念分析（Formal Concept Analysis, FCA），推衍社群網路之正、負向言論、各種經濟指數與虛擬貨幣幣值漲跌幅間的「若…則…」（if precedent then descendant）推論關係（inference relationship），建構加密貨幣價格之預測模型。為驗證分析架構之可行性，本研究爬取虛擬貨幣相關意見領袖於Twitter發表之推文、Google聲量、與影響加密貨幣價格之經濟指標，導入分析架構後，推衍預測關係。依據實證研究結果，影響比特幣價格的重要因素有比特幣交易量、黃金價格及交易量、美國股票指數、匯率、通貨膨脹率，以及Google聲量，若特定條件成立，則比特幣價格會大幅上漲或下跌；而Google聲量在投資者做出投資決策時扮演重要角色，輿論可能不會對比特幣價格的即時波動產生影響，但以整個時間序列來看，卻確切對其產生重大的影響。本研究結果可作為比特幣投資人之依據，建構之分析架構，亦可作為預測其他金融商品之用。With the advent and increasing popularity of cryptocurrencies, the use and investment of cryptocurrencies has become a new trend. However, due to the high volatility of virtual currency prices, the investment risk is extremely high. How to accurately predict virtual currency prices is of great concern. The advent and popularity of social media has changed the way most users exchange information. In addition to using search engines, people can also post or search for other people's opinions and comments through platforms such as Facebook or Twitter. Therefore, the large number of comments on social media has become the target of big data analysis, as well as for the analysis of cryptocurrencies. Previously, there were common analyses and predictions of the impact of social media comments on the stock market, but virtual currencies are a new type of investment target, and there are few scholars who have studied the impact of social media comments on the rise and fall ofBitcoin values. Therefore, this research intends to crawl tweets about cryptocurrencies from major social media platforms, conduct text sentiment analysis, and simultaneously crawl representative individual economic and macroeconomic indicators. After normalizing the data, use k-means cluster analysis to cluster each variable, and then introduce Dominance Rough Set Approach (DRSA) and Formal Concept Analysis (FCA) to derive the "if precedent then descendant" inference relationship between positive and negative comments on social networks, various economic indicators and virtual currency value fluctuations, and construct a prediction model for cryptocurrency prices.In order to verify the feasibility of the analysis framework, this research crawls tweets published by opinion leaders related to virtual currencies on Twitter, Google volume, and economicindicators that affect cryptocurrency prices, introduces them into the analysis framework, and derives predictive relationships. According to empirical research results, important factors affecting Bitcoin prices include Bitcoin trading volume, gold price and trading volume, US stock index, exchange rate, inflation rate, and Google volume. If specific conditions are met, Bitcoin prices will rise or fall sharply; Google volume plays an important role when investors make investment decisions. Public opinion may not have an impact on the instantaneous fluctuations of Bitcoin price, but over the entire time series, it does have a significant effect on it. The results of this research can be used as a basis for Bitcoin investors. The constructed analysis framework can also be used to predict other financial products.社群媒體加密貨幣情感分析k-平均演算法優勢約略集合形式概念分析Social MediaCryptocurrencySentiment Analysisk-means clusteringDominance based Rough SetFormal Concept Analysisetd