Generative and non-parametric model for real-time event detection in social networks based on textual analysis

One of the things that is followed in monitoring systems is the detection of rare events in real time among the multitude of common events in social networks. Considering the lack of recognition and unavailability of rare events, their detection is considered a challenge. In this research, a new architecture and approach based on generative adversarial network infrastructure was presented to detect common and rare events in real time. In this research, the attempt is to provide a new approach to the performance of architectures based on deep generative adversarial networks, a way to solve various problems without supervision with a semi-supervisory approach and adversarial generative infrastructure. This architecture is based on the automatic extraction and use of video input data features. The results of the equal error rate in the UCSDped1 and UCSDped2 datasets were 2.0 and 17.0, respectively, in the performance characteristic curve.