銷售預測是許多企業所注重的問題，而時尚產業中，時尚潮流快速變化、消費者需求極不穩定以及產品生命週期短，因此準確的銷售預測可避免利益損失。由於資訊科技的進步，虛擬通路的商業經營模式也隨之崛起，許多時尚產業皆增設無店鋪通路，並以實體商店結合網路商店來進行虛實通路整合的行銷策略，藉此達到更高的獲利模式。 為順應多元通路時代的來臨，本篇論文藉由時尚產業的真實數據，並考慮過去銷售資料、季節性與氣候等因素，以運用多元適應性雲形迴歸（multivariate adaptive regression splines，MARS）及一般線性迴歸以建構銷售預測之模式，來探討時尚產業中，影響虛實通路銷售的重要因素。 結果顯示，本研究所設定之預測模式及其結果可供企業的管理者作為決策之依據，並更加深入了解各重要影響因素對於未來銷售的正負影響關係，同時可取得各影響因素之轉折值，將有利於企業在複雜度較高的銷售預測中，有效的進行產品預測與行銷策略之改進。 Sales forecasting is a crucial issue that many companies pay concentrate on. Among fashion industry, due to the rapid change of fashion trend, unstable consumer demand and short product life cycle, accurate sales forecasting is able to avoid loss of benefits. As for the improvement of information technology, business model of virtual channel is also develop further. Many fashion retailer add the physical and virtual channels, combining physical store with Internet store to implement the marketing strategy with integration of physical and virtual Channel, in order to achieve higher profit models. Conform to the Multiple-channel time, the paper adopts the data of fashion industry and also considers the historical sales data , seasonality and climatic factors. And using multivariate adaptive regression splines (MARS) and general linear regression to construct the sales forecasting model. To explore the main factors that impact the sales result of physical and virtual channel in the fashion industry. The empirical results of the research, this forecasting model and conclusion can be consulted for decision making by the business manager. Also can understand the role of important factors in either positive or negative sides which affect future sales. In conclusion, the research is beneficial for the company to do products forecasting and improve marketing strategies efficiently within highly complicated sales forecasting.
