Demand Forecasting for the Full Life Cycle of New Electronic Products Based on KEM-QRGBT Model

To improve the accuracy of demand forecasting for new electronic products, especially in scenarios with limited historical data, a novel forecasting model was proposed in this study which integrated K-means based on Euclidian distance, Multi-layer perceptron algorithm, and Quantile Regression with Gradient Boosted Trees (KEM-QRGBT). The model also incorporated grid search with K-fold cross-validation to enable the adaptive selection of the optimal parameters for product data. Additionally, the KEM-QRGBT model, which can balance the intricacies of learning parameter patterns with its ability to quantify demand uncertainty, exhibited proficiency in quantifying the uncertainty inherent in demand forecasting. Using a case study from a manufacturing enterprise in Turkey, the effectiveness of the model was validated. Results demonstrate that, for new electronic products with limited historical data, the KEM-QRGBT model with adaptive parameter selection improves demand forecasting accuracy, outperforming benchmark methods, and other machine learning models. The proposed algorithm provides a strong evidence for the demand forecasting of new electronic products, particularly in cases where historical data is limited. © 2023 School of Science, IHU. All Rights Reserved.