利用網路口碑預測台北市電影票房

Using Online Word-of-Mouth to Predict Taipei Box Office Revenue

林彥佐(Yan-Tzuo Lin)

指導教授 ： 邱昭彰

元智大學/資訊學院/資訊管理學系/碩士(2012年)

摘要

電影產業是一個獲利高，風險也高的行業，因此對於電影製片商、行銷商、映演商及投資商而言，為了降低風險及不確定性，票房預測的確有其必要性。早期的研究多是以電影的特徵，如明星及導演的影響力、預算、銀幕數、上映日期等電影基本面訊息進行預測，近期的研究則是加入了口碑、評論或是新聞等文字訊息，經學者們的研究，口碑的確會影響產品的銷售及消費者的購買意願，因此做為另一項預測指標。
本研究不僅將電影特徵及口碑納入票房預測的模型中，更結合了Google搜尋透視(關鍵字熱門度查詢)，透過類神經網路(Neural Network; NN)，支援向量迴歸(Support Vector Regression; SVR)，分類迴歸樹(Classification and Regression Tree; CART)及K鄰近(K-Nearest Neighbor; KNN)並結合GA動態調整權重，分別建立模型進行預測。實驗結果發現，使用GA-KNN方法，結合電影特徵、口碑、Google 網頁搜尋熱門度的預測模型，週票房平均絕對誤差百分比(MAPE)僅1.0947，總票房MAPE僅1.1037。

關鍵字

網路口碑 ； 文字探勘 ； 電影票房

並列摘要

The film industry is often viewed as high risk but at the same time, it offers the opportunity for high profit. In order to reduce the risks and uncertainties, the box office forecasts are becoming a necessity for film producers, marketers, exhibitors and investors. In the past, research and prediction were focused on the characteristics of the film, such as the influential power of the actors and director, budget, number of screens, schedule. However in the recent studies found that factors of the reputation of the film, namely, word of mouth, film critics, news or any format of print media have been added to the accuracy of the prediction. These factors affect on product sale and purchase intent, therefore, are important prediction indications.
In this study, film characteristics and factors of reputation will be included in the box office prediction model, combined with Google Insight for research(Keyword popularity query), Neural Network (NN), Support Vector the Regression (SVR), Classification and Regression Tree (CART), and K-Nearest Neighbor (KNN). Furthermore, this model will also combine with GA dynamically adjust weights to establish the prediction model. The result has found that the model use GA-KNN algorithm that combines with film characteristics, the factors of reputation and Google web search popularity, would have only 1.0947 mean average absolute error (MAPE) weekly; and only 1.1037 for total box office.

並列關鍵字

Word-of-Mouth ； Text Mining ； Box Office

參考文獻

[2] 吳承宇(2011)，「以搜尋引擎查詢記錄為基礎之電影票房研究」，國立臺灣大學資訊管理學研究所，碩士論文。

[27] Liu, Y., Chen, Y., Lusch, R. F., Chen, H., Zimbra, D., and Zeng, S., "User-Generated content on social media: Predicting market success with online word-of-mouth," IEEE Intelligent Systems, Vol. 25, No. 1, pp. 75-78, 2010.

[4] Ahluwalia, R., "How prevalent is the negativity effect in consumer environments?" Journal of Consumer Research, Vol. 29, No.2 pp. 270-279, 2002.

[5] Askitas, N. and Zimmermann, K. F., "Google econometrics and unemployment forecasting," Applied Economics Quarterly, Vol. 55, No. 2, pp. 107-120, 2009.

[7] Basuroy, S., Chatterjee, S. and Ravid, S., "How Critical Are Critical Reviews? The Box Office Effects of Film Critics, Star Power, and Budgets," Journal of Marketing, Vol. 67, pp. 103-117, 2003.