Crime detection and crime hot spot prediction using the BI-LSTM deep learning model

Crime is defined as any act that is illegal and causes unpredictable discomfort to the common public by affecting quality of life and causing financial loss. The objective of this research work is to develop algorithms to predict crime using machine learning (ML) techniques in emotion data and predict future crime spots using crime incident data using deep learning (DL), then cross-check whether the future crime incidents match with the results of crime incidents detected. Voice-based emotion data is analysed using ML algorithms to detect crimes and crime incident data, includes audio and/or video captured from the scene of a crime with geographic coordinates, place names and timestamps are analysed using DL methods such as convolutional stacked bidirectional long short-term memory (LSTM). Crime detection using ML models provided an accuracy of 97.2% for ensemble classifiers and DL methods achieved an accuracy of 95.64% in crime hot spot forecasting.