University of Twente Student Theses
A Comparison of Deep Neural Networks for Sales Forecasting
Fitria, Silvi (2023) A Comparison of Deep Neural Networks for Sales Forecasting.
|
PDF
2MB |
| Abstract: | Accurate sales forecasting plays an integral role in Supply Chain Management (SCM) for optimizing production and allocating resources, which leads to improved profitability. Since SCM comprises numerous products in a large number of stores, this research focuses on developing a global model that can be trained in a one-step process on all time series data. The application of this forecasting uses four novel Deep Neural Network (DNN) algorithms; namely Long Short-Term Memory (LSTM), Neural Basis Expansion Analysis for Time Series (NBEATS), Temporal Convolutional Network (TCN), and Transformer. It involves a comparative analysis based on their architectures, using real-world sales data in Henkel and simulated data. The TCN model performs best in both data, exhibiting the smallest evaluation error by taking almost two hours to train the global model. Implementing the best model resulted in an average improvement of 10% in forecast accuracy compared to the current forecasting method implemented in Henkel. Keywords: Deep Neural Network, LSTM, NBEATS, Sales Forecasting, SCM, TCN, Time Series, Transformer. |
| Item Type: | Essay (Master) |
| Clients: | Henkel |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 31 mathematics, 50 technical science in general, 54 computer science |
| Programme: | Applied Mathematics MSc (60348) |
| Link to this item: | https://purl.utwente.nl/essays/95006 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page