Identifying areas vulnerable to homicide using multiple criteria analysis and spatial analysis

Highlights

• The use of maps facilitates the elicitation processes.

• Spatial analysis is used to identify significant spatial patterns.

• The maps generated by the model show the most critical vulnerable areas to homicides.

• The vulnerable areas of the neighbourhood are concentrated.

• Our model can be useful for guiding public policies.

Abstract

The decision-making process in public security is not an easy task. Several aspects must be considered, since resources are limited, whereas coverage should be extensive. Usually, preventative actions are allocated to areas that are more prone to violence, where criminal occurrences have happened in the recent past. In Brazil, such decisions are made in an ad-hoc way, considering only the knowledge of a specialist. This paper, however, aims to identify homicide vulnerability areas, taking into account the knowledge and preferences of an expert decision-maker under several criteria and data from a demographic census. Our model aggregates multiple-criteria analysis, based on Dominance-based Rough Set Approach, and spatial analysis, which consist of hot-spot analysis and local Moran's I. The model was applied in a neighbourhood of Brazil. The approach was able to highlight the problematic areas in the neighbourhood, and suggested locations where public policy and, consequently, limited resources should be allocated.

Introduction

In many areas around the world, such as Brazil, homicide rates continue to increase, while the resources to combat and prevent homicides are limited and inefficiently allocated. It is impossible to devote efforts to making interventions throughout an entire city, due to financial, human, and management limitations. Therefore, the development of new structured methodologies is an interesting approach to directing public efforts toward reducing violent deaths.

The issues of public safety and violence are often discussed because they directly affect everyone alive. Homicide cases are grave, because they involve the death of someone in the community. Violent deaths are a problem all over the world, especially in developing countries.

In 2012, 10% of the homicides in the entire world occurred in Brazil [72]. Regarding the homicide rates, homicides all over Brazil, and in each its states, exceeded the threshold for qualifying as epidemic, which according to WHO (the World Health Organization) is ten homicides per 100,000 inhabitants [68]. In 2012, the homicide rate in Brazil was 29 homicides per 100,000 inhabitants [69]. This rate, however, varies significantly across the country: among the state capitals, the rates range from 12.8 (Santa Catarina) to 64.6 (Alagoas) [69]. Moreover, the homicide trend is not the same across Brazil: while few states are experiencing a drop in homicides, this rate is increasing in other states.

Given the number of homicides, the Brazilian scenario is concerning. Several actions must be taken to reduce the number and rates of homicides to a more tolerable level. While a goal of no homicides can be considered to be unrealistic, some strategy to avert many deaths by year must be devised. In any case, there is a need for public policies that yield greater public security.

Geographic analysis of criminal occurrences is important in order to identify possible spatial patterns on which preventive actions can be targeted. According to Andresen [4], the microscale approach has been used in a number of criminal geography studies and involves identifying possible actions to be taken as countermeasures against violence [66]. Moreover, security policies in small community areas have been shown to yield better results in decreasing violence [8].

Identification of areas with the highest rates of violence or which have the greatest propensity for violence can be a first step in developing effective security policies that address this problem. Mohler [54] asserted that identifying violent areas can facilitate preventive police intervention and combat crime, and may well also optimize how police resources are allocated. However, previous studies have either used historical crime data [66] or specific data from a demographic census [58]. These approaches neglect any expert input that may enrich the decision-making process. The inclusion of experts (i.e., police officers and police security managers) strengthens the analysis and is based on knowledge of the issues and aggregating preferences. We suggest that expert opinions regarding criminality can be incorporated into the decision-making process through a structured multi-criteria approach.

This paper presents an innovative model developed inside a Geographical Information Systems (GIS) framework, based on multiple-criteria decision aid (MCDA), called GIS-MCDA, which can be used to identify areas that are more vulnerable to crime. The model integrates a spatial analysis to reveal spatial patterns, where we tested two different spatial approaches: hot spot analysis and local Moran's I. We applied the GIS-MCDA model to identify areas with high vulnerability to homicide in a neighbourhood of Recife, Brazil. The results reveal critical areas in relation to homicide vulnerability, taking into account social, economic, and demographic factors, thereby, highlighting locations where public policies can be targeted to mitigate violence in the long term.

The GIS-MCDA model is based on the Dominance-based Rough Set Approach, DRSA, proposed by Greco, Matarazzo and Slowinski [28,30] based on Rough Sets Theory [61]. Among many multiple criteria methods, we selected DRSA because it requires less cognitive effort by the decision-maker (DM). The method does not require the establishment of parameters, such as thresholds and weights [31], which are usually required by traditional MCDA methods, such as outranking based methods (ELECTRE, PROMETHEE) and additive based methods (MAUT, MAVT, AHP). The DRSA has been applied to different decision-making contexts: risk and financial performance [65], systems involving dynamic information [45], group decision-making [11,12], emergency calls [73], environmental management [51], and industrial manufacturing [36].

This paper is organized as follows: Section 2 presents a brief review of GIS-based Multiple Criteria Decision Analysis (GIS-MCDA) and Section 3 the Problem description. Section 4 describes the Methodology of the GIS-MCDA model and its Application and Results are presented in Section 5, while Section 6 brings some discussion and implications of this study. Finally, some conclusions are drawn and suggestions made for future studies in Section 7.