Behavioral Modeling of Persian Instagram Users to detect Bots

Muhammad Bazm,Masoud Asadpour

arxiv(2020)

引用 0|浏览2
暂无评分
摘要
Bots are user accounts in social media which are controlled by computer programs. Similar to many other things, they are used for both good and evil purposes. One nefarious use-case for them is to spread misinformation or biased data in the networks. There are many pieces of research being performed based on social media data and their results validity is extremely threatened by the harmful data bots spread. Consequently, effective methods and tools are required for detecting bots and then removing misleading data spread by the bots. In the present research, a method for detecting Instagram bots is proposed. There is no data set including samples of Instagram bots and genuine accounts, thus the current research has begun with gathering such a data set with respect to generality concerns such that it includes 1,000 data points in each group. The main approach is supervised machine learning and classic models are preferred compared to deep neural networks. The final model is evaluated using multiple methods starting with 10-fold cross-validation. After that, confidence in classification studies and is followed by feature importance analysis and feature behavior against the target probability computed by the model. In the end, an experiment is designed to measure the models effectiveness in an operational environment. Finally, It is strongly concluded that the model performs very well in all evaluation experiments.
更多
查看译文
关键词
persian instagram users,bots,behavioral modeling
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要