Cross market and cross source sentiment analysis driven RMB exchange rate prediction dataset

The empirical data set used in this paper is from December 17, 2017 to December 31, 2021, covering the data before and after the outbreak of COVID-19.The emotional text data in China comes from Sina Weibo, while the emotional text data in the United States comes from Twitter. Official media accounts and personal accounts at the top level are obtained based on activity and fan base, respectively, as representatives of official and market emotional information.When obtaining text data from Sina Weibo in this article, the webdriver component in Python and the BeautifulSoup crawler framework are used. Use the webdriver component to open the official webpage of Sina Weibo. In order to successfully access all Weibo posts, enter the account password to log in, and then jump to the corresponding account homepage; View the source code of the webpage, locate the data row that needs to be crawled, and set the range of pages that need to be crawled. Use the Beautiful Soup crawler framework to crawl the required Weibo and its publication time. When obtaining Twitter text data in this article, we used the Python open-source project package Scweet, which uses selenium to simulate browser access, set an account and time range, and crawl the corresponding Twitter and its release time. This article captures a total of 406877 text messages (including 145935 official media articles in the Chinese market, 26178 personal accounts in the Chinese market, 203645 official media articles in the US market, and 31119 personal accounts in the US market).Meanwhile, the time series data of the closing price of the RMB exchange rate is sourced from the Wind database.