| Popis: |
Accurate prediction of demand assists companies in responding flexibly to uncertain market conditions. However, companies face difficulties due to the wide variety of product types and the uncertainty of their scale. To solve this problem, this study proposes a hybrid model that combines K-means, ElasticNet, and Gaussian process regression (GPR). GPR effectively addresses the issue of nonlinear prediction, and its performance is excellent when the model is trained with clusters of similar data instead of training the entire dataset at once. Moreover, the model enhances its training process by extracting crucial and contributing variables for each cluster. To implement these techniques, this study utilizes K-means and ElasticNet in combination with GPR. The model is applied to a case study of a U.S. manufacturing company and is compared with other benchmarking models, including single GPR, K-means+GPR, and ElasticNet+GPR, to evaluate its performance. The hybrid model achieved the best prediction accuracy, recording a mean absolute error (MAE) of 5.57, demonstrating its potential as foundational research for constructing an efficient demand forecasting model under uncertain conditions. |
