Early Rumor Detection Based on Deep Recurrent Q-Learning

Table 1

Features of metadata.

Feature name	Description
has_url	Whether the tweet contains URL or hyperlink
urls	The number of URLs or hyperlinks contained in the content of the tweet
has_tag	Whether to use the keyword “#” to carry a topic in the content of the tweet
tags	How many topics are carried with the keyword “#” in the content of the tweet
favorite_count	The number of times the tweet was liked
retweet_count	The number of tweets reposted
verified	Is the user authenticated by real name on Twitter?
profile_use_background_image	Whether Twitter users set a background image on their homepage
default_profile_image	Whether the background image set by Twitter users on their homepage is the system default
geo_enabled	Whether users disclose their location information
is_translation_enabled	Whether the user has the translation permission
default_profile	Whether users modify the default personal information
friends_count	Number of users followed
followers_count	User attention
statuses_count	Number of tweets posted by users
description_len	User profile length
favourites_count	Cumulative number of likes of user tweets
listed_count	How many public channels are users involved
user_age	User registration period
coordinates	Does the tweet contain coordinate information
coordinates	Whether the tweet contains URL or hyperlink