Abstract
Industrial globalization has done a great impact on semiconductor supply chain. The semiconductor product demand has been rapidly raising with high diversity. Demand forecast provides vital information from supply chain to improve the quality of strategy decision making and product inventory planning. Owing to the uncertainty and variety of the products characteristics such as manufacturing lead time, semiconductor supply chain encounters difficulties when trying to build an efficient and accurate demand prediction model. Hence, this study aims to propose a novel demand forecasting model based on conditional generative adversarial network with autoencoder model to predict different products’ demand for customer demand fulfilling. For results validation, an empirical study was conducted in a leading semiconductor distributor in Taiwan. The results show the forecast accuracy improvement of proposed model in different product information scenarios.