Modeling User Behaviour Over Time from Chat Messages
As online communication among children and young adults grows in popularity, concerns about online safety are receiving vast attention. The Two Hat Security (2HS) company has a rule-based filtering system to detect malicious chat messages and identify abusive users. The proposed project will improve the predictions of the trustworthiness (trust-level) of the users which can change over time and also influences what users are allowed to say in chat messages. We will use classification and clustering algorithms to obtain predictions of the trustworthiness of the users over time using data features obtained from the chat messages. This will provide 2HS with a tool that will improve the accuracy of the trustworthiness predictions and ultimately will lead to satisfied 2HS customers as well as safe online communication for Canadian youth.