This article utilizes publicly available data to capture suspicious users in Steemit, where writers and subscribers are rewarded through posting an article and/or voting an article.
I believe that the biggest advantage of the Block-Chain technology is the disclosure of all transaction information. Steemit is built on the basis of black-chain technology in which all of the user's actions, including writing / commenting and boarding history are all publicly available. But everyone are not able to access those information easily. The purpose of this article is to provide the aforementioned information in a graphical representation so anyone can easily view/understand.
For now, I will not disclose the actual user name of the user. However, you should be aware that the block-chain technology is open to the public and any of your actions will be stored permenantly and can be easily monitored.
Period of Data Source (2017-10-07 ~ 2017-10-08)
-Total Amount of Data: 25828
-Total Number of Steem Users: 7681
Self-Voting (SV)
The above graph shows that the total number of Steemit Users who commited Self-Voting(SV) is 7681, where the descriptive statistics including average, maximum and standard deviation showed 3.36, 164 and 6.06 respectively during 2017-10-07 ~ 2017-10-08.
Top 10 Steemit Users who have commited most self-voting during 2017-10-07 ~ 2017-10-08:
Self-Voting Top 10 user's 7 day voting pattern
The above network describes the voting relationship between SV users and others where the SV users are colored with red during 2017-10-07 ~ 2017-10-08.
Self-Voting Ranked between 1~5
Self-Voting Ranked between 6 ~ 10
The above graph shows the top 10 HSV users behaviour including the total number of voting (gray), the number of SV (red), and the number of NSV (blue) during (2017-10-07 ~ 2017-10-08)
High-Self-Voter (H-SV) Steemit Users' Suspicious Scores
The Suspicious score of H-SV Steemit users is extracted with the following features:
-VP -- User's voting power
-voting_cnt -- Total number of voting during the period
-sv_voting_cnt -- Total number of self voting during the period
-nsv_voting_cnt -- Total number of voting placed for other users articles/comments during the period
-author -- Number of other users
The indicators extracted from the above features are as follows:
-sv_avg_vp -- Average voting power for self voting (=SV 총 VP합/sv_voting_cnt)
-nsv_avg_vp -- Average voting power for other users (=NSV 총 VP합/nsv_voting_cnt)
-sv_ratio -- User's self-voting rate (=sv_voting_cnt/voting_cnt * 100)
-VPA -- Average number of votes of other users (=nsv_voting_cnt/author)
The virtual indicators for compensation measurement are (USD):
-Self Rewards(SR - Self Rewards) -- The total amount of self compensatio when the user's voting reward is assumed to be $1 (=(1 * sv_avg_vp) * sv_voting_cnt)
-Not Self Rewards(NSR) -- The total amount of other users' compensation when the user's voting reward is assumed to be $1 (= (1 * nsv_avg_vp) * nsv_voting_cnt).
-Reward Per User (RPU) - Average compensation amount of other users when the user's voting reward is assumed to be $1 (= NSR / author).
Voting patterns of H-SV users during 2017-10-07 ~ 2017-10-08 including the number of voting (voting_cnt), the number of self-visits (sv_voting_cnt), the number of other users (nsv_voting_cnt), and the number of user visits (hsv_voting_cnt), Author number (authors), self-voting rate (sv_ratio), and average number of other users (VPA).
Voting patterns of H-SV users with a relatively high voting rate, SV_Ratio 9 ~ 10 - Red
The distribution of rewards for H-SV users (red) and others (blue) distributions when the reward of individual voting is assumed to be $1.
