Definição de grupos de risco de roubo ou furto, perda parcial ou total de veículos via K-means
Nenhuma Miniatura disponível
Data
2020-10-16
Autores
Prazeres, Bárbara Ivânia Rangel [UNIFESP]
Orientadores
Garcia, Raphael de Oliveira [UNIFESP]
Tipo
Trabalho de conclusão de curso de graduação
Título da Revista
ISSN da Revista
Título de Volume
Resumo
Os índices de roubo na cidade de São Paulo, entre 2018 e 2019, indicaram uma queda no número de furtos e roubos de veículos de acordo com o Ministério da Justiça (2020). Segundo a revista Veja, em 2019, a Zona Leste foi a região paulistana que reuniu seis dos dez bairros campeões de furtos de veículo. Em geral, o risco de roubo ou furto do veículo está relacionado a localização (bairro ou distrito) e consequentemente ao preço do seguro. Todavia será que isso apenas ocorre para roubo ou furto? E quanto a perda parcial ou perda total? Seria possível agrupar os distritos segundo a frequência de sinistros ou somente agrupá-los de acordo com a proximidade geográfica explicaria o risco? Com o objetivo de estudar estas frequências de sinistros considerando a geolocalização foram utilizadas as bases de dados disponibilizadas pela Superintendência de Seguros Privados (SUSEP) através do Sistema de Estatísticas de Automóveis da SUSEP (Autoseg). As bases de dados analisadas continham 60 milhões de registros (CEP, ano modelo, marca, categoria do veículo, exposição, número de sinistros de roubo ou furto, número de sinistros de perda total ou parcial). Estas bases de dados foram sumarizadas por distrito, um total de 96, onde as frequências de sinistros de roubo ou furto e perda total ou parcial foram calculadas no período de 2016 a 2018. Estas frequências foram plotadas no mapa e fora aplicado o algoritmo k-means considerando a frequência de roubo ou furto obtendo 12 clusters e ao repetir o procedimento para frequência de perda parcial ou total obtevese igualmente 12 clusters. Em geral, distritos circunvizinhos apresentaram frequência de sinistro similares, o que torna razoável agrupá-los segundo a proximidade geográfica, no entanto, ao utilizar o k-means, distritos de zonas distintas foram inseridos dentro do mesmo grupo, o que evidencia que outros aspectos influenciam no risco e devem ser considerados.
Theft rates in the city of São Paulo, between 2018 and 2019, indicated a drop in the number of vehicle thefts and thefts according to the Ministry of Justice (2020). According to Veja magazine, in 2019, the East Zone was the São Paulo region that brought together six of the ten neighborhoods that were champions of vehicle thefts. In general, the risk of theft or theft of the vehicle is related to the location (neighborhood or district) and consequently the insurance price. However, does this only occur for theft or theft? What about partial loss or total loss? Is it possible to group districts according to the frequency of claims or just grouping them according to geographical proximity already explains the risk? In order to study these claim frequencies considering the geolocation, the databases made available by the Superintendency of Private Insurance (SUSEP) through the SUSEP Automobile Statistics System (Autoseg) were used. The databases analyzed contained 60 million records (CEP, model year, make, vehicle category, exposure, number of theft or theft claims, number of claims for total or partial loss). These databases were summarized by district, a total of 96, where the frequency of theft or theft claims and total or partial loss were calculated in the period from 2016 to 2018. These frequencies were plotted on the map and the k- means considering the frequency of robbery or theft obtaining 12 clusters and when repeating the procedure for frequency of partial or total loss, 12 clusters were also obtained. In general, surrounding districts had similar frequency of claims, which makes it reasonable to group them according to geographical proximity, however, when using the k-means districts from different zones were grouped, which shows that other aspects influence the risk and should be considered.
Theft rates in the city of São Paulo, between 2018 and 2019, indicated a drop in the number of vehicle thefts and thefts according to the Ministry of Justice (2020). According to Veja magazine, in 2019, the East Zone was the São Paulo region that brought together six of the ten neighborhoods that were champions of vehicle thefts. In general, the risk of theft or theft of the vehicle is related to the location (neighborhood or district) and consequently the insurance price. However, does this only occur for theft or theft? What about partial loss or total loss? Is it possible to group districts according to the frequency of claims or just grouping them according to geographical proximity already explains the risk? In order to study these claim frequencies considering the geolocation, the databases made available by the Superintendency of Private Insurance (SUSEP) through the SUSEP Automobile Statistics System (Autoseg) were used. The databases analyzed contained 60 million records (CEP, model year, make, vehicle category, exposure, number of theft or theft claims, number of claims for total or partial loss). These databases were summarized by district, a total of 96, where the frequency of theft or theft claims and total or partial loss were calculated in the period from 2016 to 2018. These frequencies were plotted on the map and the k- means considering the frequency of robbery or theft obtaining 12 clusters and when repeating the procedure for frequency of partial or total loss, 12 clusters were also obtained. In general, surrounding districts had similar frequency of claims, which makes it reasonable to group them according to geographical proximity, however, when using the k-means districts from different zones were grouped, which shows that other aspects influence the risk and should be considered.