Optimization of Police Contingent Management Based on a Genetic Fuzzy System

Resumo

In a worldwide trend for technologies development that assist in the efficient management
of people and resources, there is a great effort by the scientific community to provide viable ways to
atach this goal. In this way, to insert in these approaches the analyses of variables that influence the
public safety administration assists in the decision making of an intelligent system. This paper proposes

a software development based on Genetic Fuzzy system to help authorities in police contingent manage-
ment in order to reduce crime in urban centers. Were selected indicators that have a greater impact on

the planning and management of public security and, from this perspective, was searched for data that
best represent the real situations and correlate with the study object. The data used in the modeling and
training of the algorithm were extracted from a database that shows the occurrence of Intentional Lethal
Crimes in Sao Lu ̃ ́ıs, Brazil. The Fuzzy C-Means (FCM) algorithm was used as a fuzzy grouping and
unsupervised learning tool to estimate the performance regions (antecedent parameters) and the rules

number of the Takagi-Sugeno (TS) inference model. The genetic algorithm is a search and optimiza-
tion technique based on the principles of genetics and natural selection. A fitness function is used in

each generation of the algorithm to gradually improve the solutions. Since the output of Takagi-Sugeno

model is the weighted sum of functions of the consequent rule, the local models parameters are opti-
mized from a fitness function. With the algorithm convergence, the answer given is a family of solutions

wich representing the best police force distribution scenarios within the optimized areas. This feature is
important for the proposed system operation, because it allows this software to function as a tool to aid
decision-making, contributing to the responsible authorities. The use of the structure of a Genetic Fuzzy
system has been shown to be efficient to the proposed problem. It can be noted as advantages of using

this method in solving problems of resource management and people the possibility of acting on contin-
uous or discrete variables, efficiency in the treatment of multiple variables, possibility of working with

numerical data, experimental and analytical functions. Computational results demonstrate the efficiency
of the methodology developed.