Sales forecasting using machine learning algorithms
Retail companies, as production systems, must use their resources efficiently and make strategic decisions to obtain growing and stable revenues, especially when market conditions are becoming more competitive and profit margins are increasingly pressured. Thus, sales forecasting is crucial to maint...
| Autores: | , |
|---|---|
| Tipo de recurso: | artículo |
| Estado: | Versión publicada |
| Fecha de publicación: | 2023 |
| País: | Brasil |
| Institución: | Sindicato das Secretárias do Estado de São Paulo (SINSESP) |
| Repositorio: | GeSec |
| Idioma: | inglés |
| OAI Identifier: | oai:ojs2.revistagesec.org.br:article/1670 |
| Acceso en línea: | https://ojs.revistagesec.org.br/secretariado/article/view/1670 |
| Access Level: | acceso abierto |
| Palabra clave: | Sales Forecast Retail Machine Learning Time Series Productive Systems Previsão de Vendas Varejo Aprendizado de Máquina Séries Temporais |
| Sumario: | Retail companies, as production systems, must use their resources efficiently and make strategic decisions to obtain growing and stable revenues, especially when market conditions are becoming more competitive and profit margins are increasingly pressured. Thus, sales forecasting is crucial to maintain competitiveness in the retail segment, but obtaining inaccurate forecasts can lead to stock shortages, causing delays in deliveries and generating customer dissatisfaction, as well as increasing inventory, increasing the cost of warehousing, forcing the “burn” of stock through promotional campaigns, directly affecting profitability. Forecasting the demand for products and services and adapting the supply chain by finding a balance has always been and will continue to be a challenge in the retail segment. This research aims to evaluate the main methods and identify the one with the greatest accuracy in sales prediction. Based on an integrative literature review (ILR), three main methods were evaluated: time series, artificial neural networks and machine learning algorithms. The results show that machine learning is more suitable in terms of accuracy, particularly when models contain exogenous and endogenous variables, in addition to allowing the identification of hidden patterns in demand that can be used to identify market trends. However, in markets with constant demands and few external interferences, its use is not justified because, for these cases, the use of time series is simpler and less costly. |
|---|