A carregar...
Publicação
Forecasting Sales of Air Conditioner Products: A Time Series and Machine Learning Approach
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 1.5 MB | Adobe PDF |
Orientador(es)
Resumo(s)
The timely installation of air conditioners is essential for enhancing customer satisfaction and
reducing operational costs. To achieve this, the company needs an accurate forecasting model
for monthly air conditioner sales volumes. This thesis explores the development of such a
model by comparing time series and machine learning approaches, specifically Facebook
Prophet and XGBoost, at the district level across Portugal. Following the CRISP-DM
methodology, the project began by understanding the business context, followed by a time
series analysis to investigate seasonal and geographical patterns. Feature engineering was
performed to incorporate time-based, weather-related, and lagged sales features, which
helped capture yearly trends and regional dynamics. Even without key external variables, such
as promotional campaign data, the proposed models surpassed the company's existing
planning approach, which relied on the previous year's values as a baseline. The results
emphasize the value of district-level forecasting in facilitating more detailed operational
planning and highlights the importance of continuous model improvement, as forecast
accuracy could be further enhanced by including additional business and market variables.
This work demonstrates the practical benefits of employing data-driven forecasting
techniques, providing a scalable foundation for aligning inventory, logistics, and workforce
planning more effectively with seasonal demand fluctuations.
Descrição
Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics, specialization in Business Analytics
Palavras-chave
Forecasting Time Series Analysis Machine Learning Models Business Intelligence SDG 8 - Decent work and economic growth