Life Cycle Prediction: A Comparison of Methods for a Lighting Products Retailer
研究比较了分段线性曲线、多项式曲线和Bass扩散曲线在预测照明产品生命周期中的效果，发现四阶多项式曲线能准确预测63%产品的生命周期，并开发了处理不平衡数据集的算法用于神经网络建模。
Product life cycle (PLC) prediction is one of the most challenging yet critically important aspects of supply chain management. Lost sales and excess inventory costs arise when there is a mismatch between demand and supply, especially at the beginning of a product’s life cycle when a new product is launched. Our proposed framework contributes to the application of decision-support systems in the prediction of PLCs of new products. In this study, we fit piecewise-linear curves, <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</i>th order polynomial curves, and Bass diffusion curves for PLC prediction and compare their effectiveness using real data from a retailer specializing in lighting products. We estimate the PLCs of 2 615 lighting products using these models and select the best-fit curve to predict their PLCs. We also develop an algorithm to address challenges posed by imbalanced datasets and apply it in neural networks for predictive modeling to determine a product’s PLC stage, demand class, and stocking decisions. The findings show that fourth-order polynomial curves can accurately predict the PLCs of 63% of the products. Bass diffusion curves emerge as the second-best performer. Our approach can be generalized to other products in other industries, and it can effectively guide end-of-life inventory decisions.