| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 946.66 KB | Adobe PDF |
Autores
Resumo(s)
Este projeto tem como objetivos principais ajustar modelos a diversos valores associados a
vendas de medicamentos e obter previsões robustas, mas adaptáveis. Para isso, foram utilizados dados
mensais sobre o número de unidades vendidas e o respetivo montante auferido, em euros. Optou-se
por focar no Alisamento Exponencial por se observar, de forma geral, a presença de uma tendência
afetada por uma forte sazonalidade nos dados. Após a revisão e o estudo mais detalhado desse
conjunto de estruturas, criou-se um programa básico a partir da linguagem de programação R que
permite a visualização dos valores e algumas métricas relevantes.
Através de funções já existentes para o propósito, o programa foi atualizado de maneira a
ajustar um modelo apropriado aos dados. Inicialmente, os parâmetros associados a essas funções
seguiram as predefinições estabelecidas. À medida que foram abordadas séries temporais com
comportamentos mais irregulares, estes foram fixados, com o objetivo de garantir que os resultados
seguissem o que era mais adequado para os dados, não apenas o que otimiza certos critérios.
Após a seleção de um modelo específico, bem como dos valores iniciais e hiperparâmetros
que o acompanham, tornou-se possível obter estimativas para diferentes horizontes temporais. Além
disso, incluiu-se no programa algumas análises dos resultados, para permitir uma avaliação das
escolhas feitas. Por fim, o código usado foi reescrito com a intenção de torná-lo mais intuitivo.
The main goal of this project is to adjust models to various values for sales of medicines, as well as to obtain robust but adaptable forecasts. To do so, monthly data on the number of units sold and the respective amount earned, in euros, was used. We chose to focus on Exponential Smoothing because we observed a trend affected by strong seasonality in most of the data. After reviewing and studying these structures in more detail, a basic program was created using the programming language R to visualise the values and some relevant metrics. Using existing functions for this purpose, the program was updated to fit a suitable model to the data. At first, the parameters associated with these functions followed established presets. As time series with more irregular behaviour were tackled, these were fixed to ensure that the results followed what was most right for the data, not just what optimised certain criteria. After selecting a type of model as well as the initial values and hyperparameters that accompany it, it became possible to obtain estimates for different time horizons. In addition, some analyses of the results were included in the program to allow an evaluation of the choices made. Finally, the code used was rewritten to make it more intuitive.
The main goal of this project is to adjust models to various values for sales of medicines, as well as to obtain robust but adaptable forecasts. To do so, monthly data on the number of units sold and the respective amount earned, in euros, was used. We chose to focus on Exponential Smoothing because we observed a trend affected by strong seasonality in most of the data. After reviewing and studying these structures in more detail, a basic program was created using the programming language R to visualise the values and some relevant metrics. Using existing functions for this purpose, the program was updated to fit a suitable model to the data. At first, the parameters associated with these functions followed established presets. As time series with more irregular behaviour were tackled, these were fixed to ensure that the results followed what was most right for the data, not just what optimised certain criteria. After selecting a type of model as well as the initial values and hyperparameters that accompany it, it became possible to obtain estimates for different time horizons. In addition, some analyses of the results were included in the program to allow an evaluation of the choices made. Finally, the code used was rewritten to make it more intuitive.
Descrição
Trabalho de Projeto de Mestrado, Estatística e Investigação Operacional , 2025, Universidade de Lisboa, Faculdade de Ciências
Palavras-chave
Série Temporal Modelação Alisamento Exponencial Previsão Programação em R Teses de mestrado - 2025
