基于泛函网络的票房预测研究与应用

发布时间：2018-01-01 23:32

本文关键词：基于泛函网络的票房预测研究与应用　出处：《重庆大学》2015年硕士论文　论文类型：学位论文

【摘要】：随着电影产业的兴起,电影资金越来越多元化。当今电影市场的主流影片主要来自民营企业的投资制作,而民营企业投资电影的最终目的必然是盈利。但是电影市场瞬息万变,电影投资也没有“黄金定律”,如何提高电影投资发行的回报率,是我们亟需解决的难题。在电影制作前期,对电影票房收入进行预测是确保电影发行投资回报,规避发行风险的重要手段。如果投资发行商能在影片制作初期对电影票房收入有较为准确的预估,就可在电影制作期间科学的投入和利用成本,在电影发行阶段合理调整营销策略,从而就能在一定程度上降低风险,保障电影的收益率。泛函网络是近年来兴起的一门新兴学科,它源自神经网络,是神经网络的一种有效拓展与延伸。泛函网络方法多样,且在众多领域都表现出较好的性能,但是目前国内外将泛函网络用于电影票房预测的研究和应用都还很少,鉴于此,本文试图将泛函网络用于电影票房预测。具体工作如下:⑴本文首先分析了我国电影市场的现状、参考了大量关于票房预测的文献,初步选出电影票房的影响因子,用Jsoup技术从网站上抓取我们需要的电影数据。对收集到的原始数据进行清洗和整合,继而将选中的参数进行归一量化处理,形成可用的训练样本集。⑵对预处理后的电影票房影响因子进行灵敏度分析,按照Morris法分析得出各参数灵敏度大小的排序,即各个参数的不确定性对模型输出结果影响大小的排序,最终确定模型的输入变量。⑶建立基于泛函网络的票房预测模型,首先,针对训练集样本进行聚类分析,使得模型训练集更加精确有效;其次,确定所建模型的拓扑结构和基函数,用高斯消元法求得模型相关参数;最后,建立电影票房预测模型,并用测试集对模型预测性能进行测试评估。⑷本文通过仿真试验验证表明:基于泛函网络的电影票房预测模型比已有的BP神经网络模型在预测票房数值时的平均相对误差更小;通过计算和比较两种模型的标准误差(root-mean-square error,RMSE),泛函网络模型的整体性能也明显优于BP神经网络。此外,基于泛函网络的票房预测模型比BP神经网络模型的结构简单、计算量小。综上所述,本课题研究能够为电影投资方和发行商提供科学合理的决策支持,具备相应的工程应用价值。同时,对于泛函网络在实际工程应用方面的推广具有一定的参考价值和指导意义。
[Abstract]:With the rise of the film industry, film funds are becoming more and more diversified. Nowadays, the mainstream films in the film market mainly come from the investment and production of private enterprises. However, the film market is changing rapidly, and there is no "golden law" in film investment, so how to improve the return rate of film investment and distribution. In the early stage of film production, the prediction of movie box office income is to ensure the return on investment in film distribution. The important means to avoid the risk of distribution. If the investment distributor can have a more accurate estimate of the box office revenue in the early stage of film production, the scientific input and utilization cost can be achieved during the production period. The rational adjustment of marketing strategy in the stage of film release can reduce the risk to a certain extent and ensure the movie yield. Functional Network is a new subject rising in recent years, which is derived from neural network. It is an effective extension and extension of neural networks. Functional networks have a variety of methods and show good performance in many fields. But at present, the research and application of Functionals Network used in box office prediction are few, in view of this. This paper attempts to use functional network to predict the box office of films. The main work is as follows: 1 this paper first analyzes the current situation of the film market in China, referring to a large number of documents on box office prediction. Select the influence factors of movie box office preliminarily, grab the movie data we need from the website with Jsoup technology, and clean and integrate the raw data collected. Then the selected parameters are normalized and the available training sample set is formed. 2. The sensitivity analysis of the influence factors of the movie box office after preprocessing is carried out. According to the analysis of Morris method, the order of the sensitivity of each parameter is obtained, that is, the order of the influence of the uncertainty of each parameter on the output result of the model. Finally, the input variable .3 of the model is determined to establish the box office prediction model based on functional network. Firstly, clustering analysis is carried out for the training set samples to make the model training set more accurate and effective. Secondly, the topological structure and basis function of the model are determined, and the relevant parameters of the model are obtained by Gao Si elimination method. Finally, the prediction model of movie box office is established. The model prediction performance is tested and evaluated by test set. 4. The simulation results show that:. The average relative error of box office prediction model based on functional network is smaller than that of BP neural network model. By calculating and comparing the standard error between the two models, root-mean-square error (RMSE). In addition, the box office prediction model based on functional network is simpler than the BP neural network model, and the calculation is less. To sum up, the performance of the functional network model is better than that of the BP neural network model. This research can provide scientific and reasonable decision support for the film investors and distributors, and has the corresponding engineering application value. It has certain reference value and guiding significance for the extension of functional network in practical engineering application.
【学位授予单位】：重庆大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP183

【相似文献】