Academia.eduAcademia.edu

Figure 2 – uploaded by Eshrag Refaee

See full PDFdownloadDownload figure
Table 2: Tweets Statistics However, a total of (14,231) tweets were excluded from the dataset due to their irrelevant content to stock market as Table 2 illustrates the tweets statistics. Total of manual tweets tagged were (118,282), and total of gold-standard (41,455) tweets were in the dataset based on their sentiment polarity. Moreover, total of (3,756) spam tweets were tagged manually including spamming IDs and keywords lists, this process prevented thousands of spammers from storing their tweets in our database. Furthermore, (58,840) tweets did not have any related meaning even it contained the research keywords due to Arabic multiple meanings of the same word.

Table 2 Tweets Statistics However, a total of (14,231) tweets were excluded from the dataset due to their irrelevant content to stock market as Table 2 illustrates the tweets statistics. Total of manual tweets tagged were (118,282), and total of gold-standard (41,455) tweets were in the dataset based on their sentiment polarity. Moreover, total of (3,756) spam tweets were tagged manually including spamming IDs and keywords lists, this process prevented thousands of spammers from storing their tweets in our database. Furthermore, (58,840) tweets did not have any related meaning even it contained the research keywords due to Arabic multiple meanings of the same word.

Related Figures (6)

Table 1: Arabic financial related tweets keywords In this study several Twitter API’s were used in parallel to crawl Arabic tweets regarding TADAWUL. However, only (10) keywords were accepted for each query, due to the limitation of crawlers which may lead to missing tweets about TADAWUL that do not contain the designed keywords. A set of search queries were generated to enhance the possibility of acquiring tweets which convey emotions, attitudes, and opinions towards a specified entity. Table 1 shows the keywords used to retrieve the Arabic financial tweets. Crawling live stream tweets was conducted between (27 August to 23" December 2015), with total of (277,453) collected tweets.
Table 1: Arabic financial related tweets keywords In this study several Twitter API’s were used in parallel to crawl Arabic tweets regarding TADAWUL. However, only (10) keywords were accepted for each query, due to the limitation of crawlers which may lead to missing tweets about TADAWUL that do not contain the designed keywords. A set of search queries were generated to enhance the possibility of acquiring tweets which convey emotions, attitudes, and opinions towards a specified entity. Table 1 shows the keywords used to retrieve the Arabic financial tweets. Crawling live stream tweets was conducted between (27 August to 23" December 2015), with total of (277,453) collected tweets.
Figure 1: Arabic tweets streaming API. Orange. The Green is positive, Red is negative, and Orange is neutral. Furthermore, the “No go icon” indicates it is spam. The “3-connected-dots icon” represents non-related tweets that do not refer to the stock market. The “location icon” refers to stock market tweets related to another country besides Saudi Arabia.
Figure 1: Arabic tweets streaming API. Orange. The Green is positive, Red is negative, and Orange is neutral. Furthermore, the “No go icon” indicates it is spam. The “3-connected-dots icon” represents non-related tweets that do not refer to the stock market. The “location icon” refers to stock market tweets related to another country besides Saudi Arabia.
We observed a phenomenon that anonymous users are selling the Twitter® followers to increase the number of followers and illustrate popularity for a Twitter account. This observed phenomenon prevented us to from storing such tweets by false popular accounts that may lead to a false analysis of this dataset. For example, a Twitter account following 180k accounts and 190k followers most probably this Twitter account user does not read the tweets of those accounts he follows due to the huge number of tweets that are generated hourly by the users. Additionally, to reduce the impact of such cases, we set a threshold of 5k accounts that each user can follow. A total of (38,432) tweets or IDs were excluded into a separate spam dataset. The Main attributes of the dataset are defined in Table 3. Table 3: Main Attributes of Data Set
We observed a phenomenon that anonymous users are selling the Twitter® followers to increase the number of followers and illustrate popularity for a Twitter account. This observed phenomenon prevented us to from storing such tweets by false popular accounts that may lead to a false analysis of this dataset. For example, a Twitter account following 180k accounts and 190k followers most probably this Twitter account user does not read the tweets of those accounts he follows due to the huge number of tweets that are generated hourly by the users. Additionally, to reduce the impact of such cases, we set a threshold of 5k accounts that each user can follow. A total of (38,432) tweets or IDs were excluded into a separate spam dataset. The Main attributes of the dataset are defined in Table 3. Table 3: Main Attributes of Data Set
Figure 3: Volume of Daily Tweets Figure 3. illustrates the Arabic tweets daily volume and amount of positive, neutral and negative tweets. These tweets were gathered from og August to 20" October 2015). Negative tweets are presented in red, Positive tweets in green and daily volume in blue color.
Figure 3: Volume of Daily Tweets Figure 3. illustrates the Arabic tweets daily volume and amount of positive, neutral and negative tweets. These tweets were gathered from og August to 20" October 2015). Negative tweets are presented in red, Positive tweets in green and daily volume in blue color.
Figure 4 : Distribution of TASI performance This study uses TADAWUL for finding the Saudi stock market trends. We collected the closing prices data for TASI through TADAWUL from (27th August to 20th October 2015). This data proved the correlations between the influential users, tweets and the stock market prices. Figure 4 shows the distribution of TASI performance, positive and negative tweets at the same time.
Figure 4 : Distribution of TASI performance This study uses TADAWUL for finding the Saudi stock market trends. We collected the closing prices data for TASI through TADAWUL from (27th August to 20th October 2015). This data proved the correlations between the influential users, tweets and the stock market prices. Figure 4 shows the distribution of TASI performance, positive and negative tweets at the same time.
Figure 5: The daily distribution of TASI performance.
Figure 5: The daily distribution of TASI performance.
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved