A carregar...
Publicação
Estudo do Comportamento do Consumidor: Análise de Associações de Consumo em Postos de Combustível e suas Lojas de Conveniência
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 8.47 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
O aumento populacional assim como o desenvolvimento da tecnologia fizeram com
que a volumetria de dados aumentasse de forma exponencial. No caso do comércio e
do retalho, o volume de dados disponível aumenta quase ao segundo, devido ao elevado
número de transações que são processadas num curto espaço de tempo. É um desafio
retirar insigths úteis destes dados, insigths esses que podem ser, por exemplo, relações
entre produtos que são adquiridos no mesmo ato de compra, através do método de MBA
(Market Basket Analysis). Estas relações, denominadas por Regras de Associação, permitem
planificar futuras estratégias de marketing, onde se podem conjugar produtos que são
associados frequentemente. Considerando o conteúdo de cada basket, é também possível
aplicar uma análise de clusters aos baskets e segmentar os tipos de visitas que uma determinada
loja recebe. Com os consumos de cada loja, é possível definir Pontos de Interesse,
na deteção de lojas com comportamentos semelhantes face a determinados produtos. É
com foco nestes tipos de análises que se definiram os objetivos desta Dissertação. São
eles 1) estabelecer relações, com o auxílio de Regras de Associação, entre os produtos adquiridos
nos diferentes baskets, utilizando um método denominado por MBA; 2) agrupar
os diferentes tipos de baskets existentes no dataset através de algoritmos de Clustering,
definindo diferentes segmentos de baskets (shopping missions); e 3) segmentar os Postos de
Combustíveis, de acordo ao seu comportamento quanto à venda de determinados produtos,
definindo Pontos de Interesse. Para o cumprimento destes objetivos foi feita, numa
fase inicial, uma revisão de literatura sobre diferentes abordagens a este tipo de problemas,
tais como RFM (Recency, Frequency and Monetary), MBA ou Análises de Clusters neste
contexto. De seguida, são descritos os todos os métodos utilizados, bem como os passos
levados em conta na etapa de Pré-Processamento e na Análise Exploratória de Dados. Os
resultados revelam associações entre alguns grupos de produtos, bem como a existência
de algumas shopping missions que vão para além da simples compra de combustível. A
segmentação dos Postos de Combustível revela a deteção de algumas semelhanças quanto
ao consumo entre postos, relativamente ao consumo de grupos específicos. O dataset analisado
é constituído por dados transacionais, disponibilizado pela GALP, e contém baskets,
as suas transações e também alguns dados de contexto geográfico e temporal.
The population growth as well as technology development have caused the volume of data to increase exponentially. In case of commerce and retail, the available volume of data increases almost by the second, due to high number of transactions that are processed in a short space of time. It’s a challenge to take insights of these volumes of data. These insights can be, for example, relationships between products that are purchased in the same purchase act, through MBA (Market Basket Analysis). This relations, called Association Rules, allow the company to plan future marketing strategies, where products that are more frequently associated can be included together. Also, considering the content of each basket, it’s possible to apply a clusters analysis on these baskets and segment all kinds of visits that a particular store receives, as well as define Points of Interest. These Points of Interest allow to detect stores with the same behaviour regarding to given products. With these kind of analysis it was possible to set the main objectives of this Dissertation. They are 1) to establish relations, with the help of Association Rules, between the purchased products in different baskets, using MBA; 2) group the different types of existing baskets in the dataset through clustering algorithms, defining different segments of baskets (shopping missions); and 3) group the Oil Stations, according to their behaviour regarding to some products sales. Firstly, to achieve these goals, it was made a literature review on different approaches to this kind of problems, like RFM (Recency, Frequency and Monetary), MBA or Clusters Analysis in this context. Then, a description of all used methods was done, as well as all steps from Pre-Processing and Exploratory Data Analysis stages. A description of used methods is done too. The results reveal the presence of associations between some products groups, as well as that there are some shopping missions beyond the simple purchase of fuel. Also, the results Oil Stations segmentation reveals that there are some similarities between stations regarding to specific product groups consumption. The analysed dataset consists of transactional data, provided by GALP, containing baskets, their transactions and also some geographic and temporal data to give some context.
The population growth as well as technology development have caused the volume of data to increase exponentially. In case of commerce and retail, the available volume of data increases almost by the second, due to high number of transactions that are processed in a short space of time. It’s a challenge to take insights of these volumes of data. These insights can be, for example, relationships between products that are purchased in the same purchase act, through MBA (Market Basket Analysis). This relations, called Association Rules, allow the company to plan future marketing strategies, where products that are more frequently associated can be included together. Also, considering the content of each basket, it’s possible to apply a clusters analysis on these baskets and segment all kinds of visits that a particular store receives, as well as define Points of Interest. These Points of Interest allow to detect stores with the same behaviour regarding to given products. With these kind of analysis it was possible to set the main objectives of this Dissertation. They are 1) to establish relations, with the help of Association Rules, between the purchased products in different baskets, using MBA; 2) group the different types of existing baskets in the dataset through clustering algorithms, defining different segments of baskets (shopping missions); and 3) group the Oil Stations, according to their behaviour regarding to some products sales. Firstly, to achieve these goals, it was made a literature review on different approaches to this kind of problems, like RFM (Recency, Frequency and Monetary), MBA or Clusters Analysis in this context. Then, a description of all used methods was done, as well as all steps from Pre-Processing and Exploratory Data Analysis stages. A description of used methods is done too. The results reveal the presence of associations between some products groups, as well as that there are some shopping missions beyond the simple purchase of fuel. Also, the results Oil Stations segmentation reveals that there are some similarities between stations regarding to specific product groups consumption. The analysed dataset consists of transactional data, provided by GALP, containing baskets, their transactions and also some geographic and temporal data to give some context.
Descrição
Palavras-chave
Big Data Clustering Machine Learning Market Basket Analysis Pontos de Interesse Regras de Associação