| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 2.49 MB | Adobe PDF |
Advisor(s)
Abstract(s)
As vendas no comércio de bens e serviços pertencem a um tipo especial de séries temporais
que normalmente contêm ambos os padrões de tendência e sazonalidade, para além de outros
aspetos, apresentando desafios para o desenvolvimento eficaz de modelos de previsão. O
objetivo principal deste trabalho consiste na comparação do desempenho de duas
metodologias de previsão na análise de séries de vendas do setor do retalho de calçado. Os
dados analisados consistem em cinco séries temporais relativas às cinco principais categorias
de calçado comercializadas pela empresa Foreva: a Bota, o Botim, a Sabrina, a Sandália, e o
Sapato. Para este estudo foram disponibilizados pela empresa dados diários das vendas destas
categorias até dezembro de 2011. Desde logo foi solicitada pela Foreva uma previsão anual
das vendas de cada uma das categorias de calçado para o ano seguinte, tendo a empresa como
objetivo a utilização dessas previsões para a especificação do número pares de calçado de cada
uma das categorias a adquirir para comercialização. Este trabalho pretendeu dar resposta a esta
pretensão da empresa confrontando as duas principais metodologias de previsão – os modelos
de espaço de estado e os modelos ARIMA. Os resultados mostram que, de um modo geral, os
modelos ARIMA têm melhor desempenho que os modelos de espaço de estado na previsão
de séries de vendas do setor do retalho de calçado.
Retail sales of goods and services belong to a special type of time series that typically contain both trend and seasonality patterns, in addition to other aspects, presenting challenges for the effective development of prediction models. The main objective of this work consists in comparing the performance of two methods of prediction on analysis of industry sales in the retail of footwear. The data analyzed consist of five series related to the five main categories of footwear marketed by the company Foreva: the Boot, the Bootie, the Flat, the Sandal and the Shoe. For this study were made available by the company daily data from sales of these categories until December 2011. Since then was requested by Foreva a forecast annual sales for each category of footwear for the following year, and the company aimed to use these predictions to the specification of the number of pairs of shoes each category to acquire for marketing. This work aimed to address this claim of the company contrasting the two main methods of forecasting - the state-space models and ARIMA models. The results show that, in general, ARIMA models have better performance than the state-space models in time series forecasting industry sales of retail footwear.
Retail sales of goods and services belong to a special type of time series that typically contain both trend and seasonality patterns, in addition to other aspects, presenting challenges for the effective development of prediction models. The main objective of this work consists in comparing the performance of two methods of prediction on analysis of industry sales in the retail of footwear. The data analyzed consist of five series related to the five main categories of footwear marketed by the company Foreva: the Boot, the Bootie, the Flat, the Sandal and the Shoe. For this study were made available by the company daily data from sales of these categories until December 2011. Since then was requested by Foreva a forecast annual sales for each category of footwear for the following year, and the company aimed to use these predictions to the specification of the number of pairs of shoes each category to acquire for marketing. This work aimed to address this claim of the company contrasting the two main methods of forecasting - the state-space models and ARIMA models. The results show that, in general, ARIMA models have better performance than the state-space models in time series forecasting industry sales of retail footwear.
Description
Dissertação apresentada ao Instituto Politécnico do Porto para obtenção do Grau de Mestre em Logística
Orientada por: Professora Doutora Patrícia Alexandra Gregório Ramos
Keywords
Modelos de espaço de estado inovativos Calçado Previsão Modelos ARIMA Alisamento exponencial Comércio a retalho Vendas Innovations state space models Shoes Forecasting ARIMA models Exponential smoothing Retail sales