近年來藥局逐漸成為患者或消費者最重要的取藥途徑,且隨著消費習性的改變,連鎖藥局已是藥品零售業者主要的發展型態,而根據資料顯示民眾對於藥品的需求逐年增加,為了提高利潤業者必須改善存貨管理品質,其中需求預測扮演著重要的角色,然而由於藥品種類繁多且特性複雜,因此準確的預測十分困難,於是本研究提出一套適用於不同特性藥品的需求預測系統,藉由分析藥品需求歷史資料的充足性、隨機性以及季節性將藥品分為四大類:新產品、隨機性長期型產品、無季節性長期型產品以及季節性長期型產品,其中對於新產品以及隨機性長期型產品以灰色預測法進行預測,而無季節性長期型產品以及季節性長期型產品則以時間序列法進行預測,包含移動平均法、指數平滑法以及自我迴歸整合移動平均,並用分解法處理藥品需求的季節性,此外,本研究也探討新資料加入時的再預測,透過預測模型的適用性評估以做為延續或重建模型之依據,同時也進一步評估歷史資料的適用性以去除不適用的資料。最後,透過實際資料對本研究所提出的預測系統進行驗證與分析,結果顯示預測系統能夠適時依據需求資料的特性改變其產品分類並選用適合模型之機制,確實有助於預測準確度的提升並且大幅減少預測所需的時間。
Forecasting enhances the efficiency and effectiveness in decision-making. Through demand forecasting, retailers not only handle demand uncertainties but also improve inventory management. However, demand forecasting for drugs is more complicated due to the various types and features. In this paper, we propose a demand forecasting system for regional pharmacy chain stores. Drugs are categorized into four types based on sufficiency, randomness and seasonality, namely, new products, random long-term products, non-seasonal long-term products, and seasonal long-term products. Grey forecasting method is applied to forecast new products and random long-term products. We also apply time-series methods, which include moving average methods, exponential smoothing methods, and ARIMA, to forecast non-seasonal long-term products and seasonal long-term products. Then, the seasonality of historical data is analyzed through decomposition method. Moreover, we discuss whether the model and parameters should be reconstructed with new data. The suitability of data is simultaneously discussed in the system. Finally, we verify our forecasting system with real data. Results indicate that the proposed forecasting system can determine a suitable model for predicting demand accurately.