Publicação
Samsung field lab Voids: demand forecasting for consumer electronic goods - a transformer-based deep learning approach
| Resumo: | VOIDS is a data science student entrepreneurial project that aims to integrate marketing into demand planning, helping companies to achieve the most accurate way of planning and shaping future demand. The following work applies this vision in a lean agile start-up framework, implementing state-of-the-art deep learning time series forecasting techniques for the global consumer electronic brand Samsung. This work implements Google AI’s Temporal Fusion Transformer to forecast demand. Novel algorithms then optimize forecasts related to Samsung’s operational processes and context and benchmark against a customized measure of accuracy, achieving a 11,4-pp. absolute increase. Further, applicability to demand shaping is discussed. |
|---|---|
| Autores principais: | Wandersleb, Tobias Theodor |
| Assunto: | Marketing Forecasting Machine learning Python Business analytics Digital transformation Deep learning Demand planning Consumer electronics Lean startup |
| Ano: | 2022 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso embargado |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | VOIDS is a data science student entrepreneurial project that aims to integrate marketing into demand planning, helping companies to achieve the most accurate way of planning and shaping future demand. The following work applies this vision in a lean agile start-up framework, implementing state-of-the-art deep learning time series forecasting techniques for the global consumer electronic brand Samsung. This work implements Google AI’s Temporal Fusion Transformer to forecast demand. Novel algorithms then optimize forecasts related to Samsung’s operational processes and context and benchmark against a customized measure of accuracy, achieving a 11,4-pp. absolute increase. Further, applicability to demand shaping is discussed. |
|---|