PhD Defense of Nasir Mahmood (2013-PhD-CS-03)

Posted On Tuesday, August 1, 2023


Abstract: In recent years, the use of social media has increased exponentially which has produced lot of data. This data is big motivation for researcher to analyze the information provided in the posts. These information may lead toward very useful knowledge for operational and strategic policy making. The human behavior can be predicated from these social media posts. The offensive human behavior predication is helpful in tracking the potential criminals and incidents. The regular occurrence of violent attacks, public protection has become a rising concern, with global crime incidents viewing a year-on-year growth trend. Developing a comprehensive and instantaneous alarming system can efficiently minimize or remove public safety issues while also ensuring the safety of citizens and their property. Identifying human behavior is a primary consideration in monitoring as soon as possible before the complexity of human behavior reco nition occurs, which is indispensable for crime prevention. In surveillance, automatic detection and recognizing of offensive behaviors facilitate immediately addressing of risks. Machine learning (ML) can be used for modeling and prediction of negative human behavior from social media contents and existing reports. The model can be developed which classify the offensive hum in behavior and also types of offensive behavior. In this research work, enhanced federated learning fused model is developed to predict human behavior and classify the offensive human behavior that may assist in identifying any criminal activity in real-time. The proposed approach's simulation results show better accuracy and miss rate results.