Abstract: Currently, the film quality assessment has focused on various fields such as complex networks, communication, economics and literature. Among them, for modelling and obtaining the properties of collaborative relationship among actors, using the complex network as a research tool has become a major approach. However, as the real world is becoming more complex and diversified, actor collaboration methods and relationship tend to become more complicated. At the same time, based on complex networks' modelling methods, it is difficult to fully represent some of the unique characteristics that the system has. Conversely, in real networks, which are modelled using the hypernetwork theory, it has been able to help researchers to better understand the complexity of real systems. Moreover, it must be noted that most of scholars have conducted empirical studies from the perspective of film production, while few investigate its linkage with the participating actors. In fact, the influence of actors is not only a key factor in determining the quality of a film, but also an important reference for casting the film. In other words, in the process of film production, the selection of influential actors and actresses helps to improve the quality of films, thus promoting the development of the film industry. Therefore, under the background of the Internet, this paper explores how the quality of a film can be improved through the selection of actors.
Based on the hypergraph theory, this paper takes films as hyperedges and actors as nodes, establishes an actor hypernetwork and obtains the topological properties of this hypernetwork. Moreover, to better analyze the influence of actors under the background of the Internet, this paper proposes three actor influence analysis models based on the hypernetwork structure, weights and the combination of both. 1)Firstly, we us the actor hypernetwork model to calculate topological metrics of nodes in a hypernetwork. Then, we use it to construct an actor influence model based on a hypernetwork structure. Finally, the model is used to estimate actor influence. 2)Given the limitations of the physical structure of the hypernetwork topology metrics, i.e., they do not accurately reflect the reality of actors' personal values, and therefore, this paper counts the actor's awards, the number of followers on social platforms and stylistic positioning. Based on empirical data, this paper constructs an alternative model of actor influence analysis to further analyze the reality of actor. 3)Based on the two models, considering the topological indicators of the hypernetwork and realistic influencing factors, this paper proposes a comprehensive actor influence analysis model, which provides a comprehensive assessment of actor influence. Sample data are collected from the official platform (https://www.maoyan.com/), which covers data periods from January 2015 to December 2019, totaling 551 domestic films released in the mainland and 2,157 actors. In addition, several categories of authoritative awards are included: The Golden Statue Awards (GSA), the Golden Rooster Awards (GRA), the Golden Horse Awards (GHA), the Hundred Flowers Awards (HFA), and the Ornamental Column Awards (OCA). Furthermore, the number of actors' fans is counted on the official platform (https://weibo.com/).
Using the three actor influence analysis models to analyze and compare actors' influence, the research results reveal that the node degree and node hyperdegree of actor-weighted hypernetworks follow a power law distribution. In addition, the hypernetwork has large aggregation coefficients and small average path lengths, i.e., it has small-world properties. Specifically, the hypernetwork structure and weights-based combination can reflect the actual influence of actors more comprehensively and objectively. The results calculated by the first model are deviated from the reality perception. The analysis result of the second model is that actors such as “Andy Lau” and “Zhang Ziyi” have strong influence, which is in line with the reality. However, there has been a lack of actors' personal value in recent years, and the result is not accurate enough. In contrast, the third model considers the role of an actor's personal value and audience recognition in his influence in recent years, which is a more objective and comprehensive reflection of the actual influence of actors in the film industry. And this paper proposes that the influence analysis method can be used not only for the prediction of actors' influence, thus providing a certain reference for film casting, but also for the influence analysis of other industries. Finally, it is also possible to study the importance of actors from the perspective of a cooperative game and further identify the most compatible way of assessing the characteristics in the actor's industry.

Key words: hypergraph, actor hypernetwork, scale-free, actor influence, analysis model of actor influence

摘要： 演员是确保一部影片成功的关键因素,在电影制作过程中,选用有影响力的演员有助于提高影片质量、推动电影产业的发展。本文以猫眼官方平台2015—2019年在内地上映的551部国产影片,2157个演员为数据样本,将电影抽象为超边,演员抽象为节点,构建了演员超网络模型,并获得了该超网络的拓扑性质。为了更好的分析互联网背景下的演员影响力,本文提出了基于超网络结构、权重及超网络结构与权重结合的三种演员影响力分析模型,并利用三种影响力分析模型对演员的影响力进行了对比分析。结果表明,基于超网络结构和权重结合的影响力分析模型能更全面、更客观地反映演员的实际影响力。此外,本文提出影响力分析方法不仅可用于演员影响力的预测,也可用于其他行业的影响力分析。

关键词: 超图, 演员超网络, 无标度, 演员影响力, 演员影响力分析模型