在科技的進步下,近年來電影產業發展快速,票房在全球逐年突破新紀錄,而市場上任何一部電影上映之後,最終以「高口碑高票房」、「高口碑低票房」、「低口碑高票房」、「低口碑低票房」,這四種結局找到歸宿,過去研究中沒有一個分類標準來區分這四種結局,故本研究利用電影相關特性來做為區分票房與口碑之重要變數,並利用鑑別分析(LDA)、多元適應性雲形迴歸(MARS)、分類回歸樹(CART)等分類方法,探討影響分類的決定因素以及其影響程度,以及方法適用性。 結果顯示,本研究使用多元適應性雲形迴歸,在分類準確率明顯優於另外兩種分類方法,CART整體準確率雖較差,但透過CART分類規則,可以找出隱藏在資料中的重要影響變數,可提供電影產業相關業者,利用現有的資源與網路資源,掌握電影票房銷售與口碑可能落點,並規劃應對的行銷方案,創造出更大的收益結果。
Because technology rapidly improves , the film industry is flourishing in these years , which creates new record of box office year by year .Then it can be divided into several categories of result in market ,which are「high WOM High box office」,「high WOM low box office」,「low WOM High box office」and「low WOM low box office」.There was no standard to distinguish these four kinds of result before .Consequently , the features of movies are utilized to variable reference of box office . This study used LDA , MARS and CART to select their inflential,and their relative importance.The result shows that MARS is used in this research ,which is better than the other two classification models.CART is less accurate in general ,but it can find out the critical variables in the movie classification ,which can be provided to film industries to grasp the market with current information and internet resource .Therefore,film firm can make appropriate sales strategy and gain more benefits.