在現今快速變化及高度競爭的產業環境中,一種嶄新的商業模式-「快速時尚」在服飾產業帶來一波新的革命。在時尚產業裡,缺乏歷史數據、時尚趨勢不斷變化及產品需求不確定性情形下,準確的銷售預測是一個重要且具有挑戰性的問題。本研究整合K-Means集群技術與極限學習機(extreme learning machine, ELM)及支援向量迴歸(support vector regression, SVR)分別建構以集群為基礎KM-ELM及KM-SVR之時尚產業銷售預測模式,並根據日本時尚產業個案公司的實體店面與無店鋪通路銷售資料進行實證分析。研究結果顯示,KM-ELM及KM-SVR模式均優於ELM和SVR模式而有較高的預測準確度,表示整合集群技術有助於改善預測表現。此外,無論是在實體店面或是無店鋪通路銷售預測上,KM-ELM模式皆有良好的結果;其相較於其他預測模式,可為時尚產業界之最適銷售預測方法。
In fast fashion industry, accurate sales forecasting is essential and challenging, because of ever-changing fashion trends, insufficient historical data, and uncertainty in demands. This study propose clustering-based sales forecasting models which inte-grate K-Means and either of extreme learning machine (ELM) and support vector re-gression (SVR), namely KM-ELM and KM-SVR. The multiple-channel retailers of Japanese fashion company is selected as a case study in the research to do the empiri-cal analysis. The results showed that KM-ELM and KM-SVR provide better forecast-ing accuracy than ELM and SVR, and the results also presented that K-means is help-ful for improving the forecasting performance. Comparing with other forecasting models, KM-ELM performs better forecasting accuracy in multiple-channel retailers which can be seen as the best model of sales forecasting in fashion industry.
為了持續優化網站功能與使用者體驗,本網站將Cookies分析技術用於網站營運、分析和個人化服務之目的。
若您繼續瀏覽本網站,即表示您同意本網站使用Cookies。