新上市產品銷售預測為許多公司所重視的課題，銷售預測通常都需要參考過去長時間的歷史資料後，才能精準預測，但新品往往無長時間歷史資料可供使用，因此本篇論文著重於利用現有的資料，於新品上市初期預測銷量。 本篇論文使用電影票房預測為案例，並以搜尋引擎做為資料取得的來源，探討電影的固定效用和競爭與電影的網路口碑或不同的季節性效果與電影票房的影響，並針對不同預測票房的組合去進行比較，最後再進行票房預測後的分析。為使電影票房的預測更加精準，利用一般線性迴歸、非線性迴歸及多元適應性雲形迴歸(MARS)等，分析數個銷售預測的模型，而獲得最佳預測結果。 本模式之結果將有助於未來企業面對新品上市時，可於短時間內利用現有的資源，於短時間內預測出新產品銷售，或快速進行上市策略的改善及調整。 New product sales forecasting is a crucial task to many innovative companies. Conventionally, the accuracy of sales forecasting is conditional on long-term, sufficient data of sales history. However, the sales information involved with newly launched products is very limited, unavailable or inaccessible. Since virtually all sorts of products are seasonal, the main concept of this study is to propose a seasonal sales forecasting model in the setting of box office for U.S. motion picture market using online open access data, such as critics, comments and ratings, which not only exhibit original interests of different stakeholders in a specific movie but also are available at both pre- and post-released phases. The results of this proposed nonlinear model with desired accuracy manifested that the influence influence of online word of mouth and seasonality significant on box office in the U.S market, contrary to those of the movie characteristics and competitions both very insubstantial.
