| 摘要: | 隨著全球市場迅速變遷及科技進步，銷售預測在企業財務風險管理、庫存控制與生產製造規劃中扮演關鍵角色。精準的銷售預測不僅能提升企業成本效益與客戶滿意度，還可減少生產庫存風險，進而增強企業競爭力與財務穩定。因此，有效的銷售預測對於應對市場需求變化及優化庫存與生產管理至關重要。
近年來，由於電子產品的廣泛運用及居家辦公模式的興起，消費者對於護眼閱讀及智慧照明產品的需求明顯增加。在選購如電腦螢幕等電子產品時，消費者往往也會同步考慮購買螢幕護眼照明裝置。面對此趨勢，個案公司建立安全庫存水位以應對市場需求的波動，確保產品供應穩定性。因此，時間序列模型應用於螢幕掛燈的短期銷售預測顯得尤為重要。然而，除了季節性變化和市場趨勢的影響，同時需要考量如疫情等特殊事件的衝擊，使得短期銷售預測充滿挑戰。
本研究對市場A、B中的三種螢幕掛燈ScreenLight A、B、C，應用Prophet、SARIMA及Holt-Winters三種時間序列模型，訓練區間分為1年、1.5年及2年來進行未來三個月的短期銷售預測，以評估這些模型在處理季節性變化及突發事件時的效能。研究結果顯示，由於Prophet模型在對季節性變化和特殊事件的快速響應方面具有優異的適應力，使其在三種模型中展示了較高的預測精確性。相較之下，SARIMA和Holt-Winters模型雖在特定條件下表現良好，但在更廣泛的應用場景中適應性相對較弱。
本研究展示了短期銷售預測在螢幕掛燈市場中的重要性，同時為相關企業提供了具體且可行的策略，以增強其面對市場不確定性及外部衝擊時的快速反應能力。
;With the rapid changes in the global market and advances in technology, sales forecasting plays a crucial role in corporate financial risk management, inventory control, and production planning. Accurate sales forecasting not only enhances corporate cost-effectiveness and customer satisfaction but also reduces the risks of overproduction and insufficient inventory, thereby boosting competitive advantage and financial stability. Thus, effective sales forecasting is essential for responding to market demand fluctuations and optimizing inventory and production management.
In recent years, the widespread use of electronic products and the rise of the work-from-home model have significantly increased consumer demand for eye-protective reading and smart lighting products. When purchasing electronic products such as computer monitors, consumers often consider acquiring screen protective lighting devices simultaneously. In response to this trend, the case company has established safe inventory levels to stabilize product supply in the face of market demand fluctuations. Therefore, the application of time series algorithms for short-term sales forecasting of screen lighting lamps becomes particularly important. However, the forecasting process is challenging due to the impact of seasonal variations, market trends, and special events like pandemics.
This study applies three time series algorithms—Prophet, SARIMA, and Holt-Winters—to forecast the short-term sales for the next three months of three types of screen lighting lamps, ScreenLight A, B, and C, in markets A and B. The training intervals are set at 1 year, 1.5 years, and 2 years, to assess these models′ efficacy in handling seasonal changes and sudden events. The results show that the Prophet model, due to its excellent adaptability in responding quickly to seasonal changes and special events, demonstrated higher forecasting accuracy among the three algorithms. In contrast, while the SARIMA and Holt-Winters models performed well under specific conditions, their adaptability in broader application scenarios was relatively weaker.
This research underscores the importance of short-term sales forecasting in the screen lighting lamp market and provides specific and feasible strategies for related businesses to enhance their rapid response capabilities in the face of market uncertainties and external shocks. |
