Predicting criminal activities analyzing video signal using machine learning

Abstract

Criminology is a method that is used to perceive wrongdoing and criminal qualities. The crooks and the wrongdoing occasion likelihood can be overviewed with the help of criminology frameworks. Video analysis and machine learning tasks have been moving from inferring the present state to predicting the future state. Law enforcement agencies can work e ectively and respond faster if they have better knowledge about crime patterns in di erent geological points of a city. In this thesis, we proposed a system to predict criminal activities by using di erent neural networks and machine learning algorithms and approaches. The target of this proposed model is to break down dataset which comprise of various violations and anticipating the kind of crimes which may occur in future relying on di erent conditions. Contrasted with other existing models, we utilized another neural systems calculation called fastGRNN which is quicker and powerful. The experimentation is conducted on various datasets. Binary classi er, CNN, GRNN, Decision Tree, Support Vector Machine were used during experimentation. By implementing these algorithms, we came down to an accuracy of 89%.