Crime Analysis and Forecasting using Twitter Data in the Indian Context

Since the late 1990s, social media has added more features and users. Due to the rise of social media, blogs and posts by common people are now a part of mainstream journalism. Twitter is a place where people can share their ideas about culture, society, the economy, and politics. India's large population and rising crime rate make it hard for law enforcement to find and stop illegal activities. This article shows the use of Twitter data to analyse, forecast, and visualise criminal activity using statistical and machine learning models and geospatial visualisation techniques. This helps law enforcement agencies make the best use of their limited resources and put them in the right places. The research aims to present a spatial and temporal picture of crime in India and is split into three parts: Classification, Visualisation, and Forecasting. Crime tweets are identified using a hashtag query argument in the tweepy python package's search_tweets function, followed by substring-keyword classification. The visualisation uses gmaps and bokeh python packages for geospatial and matplotlib for analytical applications. The forecasting portion compares AR, ARIMA, and LSTM to determine the best model for time series forecasting of crime tweet count.