Feature extraction and development of artificial intelligence based techniques for detection of fake accounts and account groups in social media

With the development of technology has developed in social networks. Access to social networks can be provided at home, at work, at any moment of life. Not only for communication and socialization, but also for many activities such as shopping and access to information. Twitter is in these social networks. Some malicious accounts on Twitter are creating false information and agenda. This is one of the main problems in social networks. Therefore, the importance of detecting malicious accounts is increasing. In this thesis, it is envisaged to classify harmful and harmless accounts in order to keep people away from this false information. It is aimed to solve the problem by categorizing the data set with advanced machine learning algorithms. For this purpose, feature extraction has been made in human and bot accounts. The features obtained were trained with different machine learning algorithms and their performances were compared. In addition, the effects of feature enhancement and multiple machine learning methods on performance were investigated. Multiple sequence alignment method was used for the behavior of bot groups. Different results were obtained by applying machine learning algorithms to different data sets. It has been observed that the results change when the attributes used change. The results obtained were compared with each other. As a result, 99% performance rate was obtained.