Crime Classification Using GRU, CNN And Autoencoder Techniques
Main Article Content
Abstract
An activity that violates the law is considered a crime. It is harmful to society to understand crime in order to avoid criminal action. The various research works are performed regarding crime prediction. Many researchers’ work based on San Francisco Dataset and analyzed the common crime pattern. They have failed to collect real time dataset for the particular region and classify the crime against women. In this work we have collected real time dataset and classify the crime based on the activity of offender that happened in Thoothukudi district. This helps to find specific crime like crime against women and child and helps to take necessary action to reduce the crime. Using this we can analyse the reason and history of offender. For that we have used One Dimensional Convolutional Neural Network (1DCNN), Gated Recurrent Unit(GRU) and Autoencoder deep learning techniques to improve the performance of crime classification. For our dataset Autoencoder yields high accuracy, sensitivity, F1-score and MCC. GRU provides better precision over the 1DCNN and Autoencoder.