Spatial visualization on patterns of disaggregate robberies

The problem of aggregating large samples of criminal data in visual representations, as often observed in many studies using geographic information systems and optimization tools to perform social assessments and design spatial patterns, is discussed in this work. A compensation bias in the correlation measure of the spatial association can be found in such types of big data aggregations, which may jeopardize the entire analysis and the conclusions from the results. In this work, a big dataset of robbery incidents recorded from the years 2013 through 2016 in Recife, one of the most important Brazilian capitals, is decomposed into 9 small sets of specific robberies, namely, Larceny, Armed Robbery, Group Stealing, Motor Vehicles Thefts, Burglary, Commercial Burglary, Saidinha de Banco (Saucy Bank), Motor Vehicle Robbery (Carjacking) and Arrastão (Flash Robbery). More accurate measures for the spatial autocorrelation can be derived from the individual incidences as proposed in this work. The visualization of optimized hot spots and cold spots of crime based on these autocorrelation measures besides enable rapid actions where crime concentrates, they have the property to design spatial patterns that can be associated with environmental, social and economic factors to support more efficient decision making on the allocation of public safety resources.