电影票房作为电影行业最为主要的收入来源，研究其影响因素并对其进行预测，有利于电影行业的发展和投资者做出正确投资决策。本文选择了2011年至2017年，除动画电影外的100部国产电影，在创作、市场和营销力度三个方面共选择18个指标构建指标体系。首先用逐步多元回归模型对电影票房影响因素进行分析，然后选择10部电影进行预测，通过多元回归、BP神经网络、随机森林和SVR模型构建票房预测模型，并比较几个模型的拟合程度和稳定性，发现随机森林模型的预测效果最好，因此用随机森林模型作为预测电影票房的模型更加合适。
More and more people and investors focus on film industry in the context of promoting the development of film industry. Studying on the influence factors of box office and predicting box office can benefit the development of film industry and can be the base for investors to make right decisions. This article selects top 100 Chinese mainland domestic films since 2011, which excludes animated films. We select and define 18 indexes to build the indexation system of influence factors of revenues of releasing films. We use stepwise multiple regression method to analyze the influence factors of box office. From the result, the market factor has the best effect. We can predict box office from multiplied regression, Neural Networks, Random forest and SVR model. We can find the best fitting model by comparing these four models. We can find that the random forest model is the best model.
Hennig-Thurau, T., Houston, M.B. and Walsh, G. (2007) Determinants of Motion Picture Box Office and Profitability: an Interrelationship Approach. Review of Managerial Science, 1, 65-92.
