Mining the Relationship between Crimes, Weather and Tweets

Abstract

This research project attempts to correlate crime rates in Orlando, Florida to Orlando's weather and Twitter presence. The central dataset of interest details the crime incidents in Orlando, Florida as reported daily by the Orlando Police Department. This dataset gives the dates, categories (e.g. theft, aggravated assault, etc.), and latitude and longitude of each reported crime incident. Using a Twitter developer account, Tweets pertaining to crime are downloaded from the greater Orlando area. Tweets are filtered by the following indexed keywords: "crime", "drugs", "narcotics", "weapons", "assault", "theft", "robbery", "murder", and "larceny." Additionally, Orlando's daily weather data is collected from the National Oceanic and Atmospheric Administration. Using measures of similarity, it is discovered that crime in Orlando is concentrated most closely near Orlando's downtown center. Using regression, moderate correlations are drawn between the rates of crime and the posting of crime-related Tweets. Lastly, chi-square tests are used to show the effect of weather on crime. High crime rates are associated with average daily temperatures above 60°F. Low crime rates are associated with days with precipitation.