以搜尋引擎查詢紀錄為基礎之電影票房研究

Abstract

銷售預測往往影響著組織的管理與策略的制定，準確的銷售預測能使企業能夠保持競爭優勢，而銷售預測的誤差會對企業造成相當大的影響。在資訊科技發達及 WEB 2.0 興盛的年代，數位口碑及網路行為能表達消費者的興趣及趨勢，過去口碑與銷售量影響的研究均針對商品的數位口碑進行分析，從中取得口碑與銷售量的關係來預測銷售量，本研究則強調以消費者主動的角度，利用群眾智慧的力量以及數位口碑來讓銷售量的預測更為準確，我們利用搜尋記錄(Query Log)代表消費者的興趣及主動性，透過 Google 的 Google Insights for Search 服務來取得使用搜尋記錄，並將查詢值結合了時間序列(Time Series)的貝斯模型(Bass Model)來預測電影票房，利用過去三年的電影資料，我們的研究證明了搜尋記錄的確有助於電影票房的銷售預測，而研究中也驗證了搜尋記錄在電影票房銷售預測中的重要性。

Sales forecasting is essential to decisions making and processes planning. Any error of sales forecasting will affect the competitive advantage of corporations. In the era of Web2.0, word-of-mouth and network behaviors are representative of consumer interests. In this study, we study word-of-mouth effect and the power of wisdom of crowds for sales forecasts. We also incorporate the query logs of search engines into a time series based Bass Model to construct a sales prediction model which produce accurate sales forecasting. Evaluations based on a movie dataset demonstrate that the query logs of search engines are effective in box office predictions.