Evaluating Trustworthiness of Twitter Users: Addressing Data Drift and Domain Relevance Through Novel Dataset Creation and Feature Analysis

This research address the challenge on data drift and domain relevance on the use of machine learning for twitter user trustworthiness classification. Initially, a Finland politician twitter user dataset [2] achieved a 97% training accuracy but dropped to 65 % when tested on Indonesian twitter user data, indicating significant data drift. A novel Indonesian dataset was achieved to 100% training accuracy, 94% validation accuracy and 95 % testing accuracy on politician twitter user. The model suffered when tested on another domain with the 79% of the avarage accuracy. The feature importance and SHAP values revealed key features influencing trustworthiness, such as content score, like ratio, and friends count. This study highlights the need for tailored datasets and domain-specific approaches to enhance model reliability.