_edited.png)
Authors: Tracy Nguyen, Jerry Jiang, Samah Basit
Claremont University, 2021
Introduction
The Marvel Cinematic Universe is an American media franchise that has gained a substantial fanbase globally. It is a shared universe center on a series of superhero films, independently produced by Marvel Studios and based on characters that appear in Marvel Comics throughout its history. The Marvel Cinematic Universe shares many similarities and draws insights from the original Marvel Universe in comic books by the crossing over plots, settings, cast, and characters. As most of the existing characters in the Marvel Cinematic Universe have precedence in the Marvel Comics Universe, Marvel has followed a pattern to select the characters in the cinematic universe based on their popularity, impact, and relations in the original comics universe.
Heroes/villains' partners have been a central element to Marvel's success. They enrich the storylines and complicate the plots to make the stories more intriguing to the general viewers. Additionally, the collaborations between the villains and heroes mainly have made the Marvel Universe storyline unpredictable, creating an additional level of excitement. In this social network, the nodes represent the characters and the edges that linked pairs of nodes, denoting different collaboration types. This paper aims to analyze an artificial collaboration network among the villains and heroes to measure the transitivity of the overall Marvel network and identify if certain characters have a strong centrality that would provide insights on Marvel's future addition of characters to its overall cinematic world.
Research questions
There are two research questions the paper intending to address.
1. What heroes/villains have the highest connection?
2. What recommendations can/will be made for future introductions in movies and comics based on the connections among the heroes/villains.
Background
The paper's literature is based on Marvel Universe Looks Almost Like a Real Social Network, by Alberich, Miro-Julia, and Rossello. An important fact that we were able to learn from the article was that the Marvel Universe is a purely artificial social network whose nodes correspond to entities that are created by Marvel and links that writers have completed for over forty years. Alberich, Miro-Julia, and Rossello found that although the Marvel Universe captures certain human relations and real-life collaborations, the network remains certain artificial traces. They conclude that "the artificiality of the Marvel Universe network lies mainly on the distributions of edges in the bipartite graph which yields it, because the relationship between the Marel Universe network's data and those of its null random model is similar to that of real-life collaboration networks' data and their corresponding null random models" (Alberich, Miro-Juilia & Rossello, 2002). Therefore, the result from this paper would not be viewed as a pure social-network analysis; instead, it would serve more as a supplementary study to provide suggestions on Marvel's particular business case.
The leading social network theories this paper incorporates are homophily, transitivity, and centrality. The homophily theory focuses on the idea that characters are more likely to create connections, relationships, or associations if they have common attributes. This theory makes it easy to understand why certain collaborations exist between heroes, villains, and antiheroes and how nodes are drawn together because of these common attributes.
Since the network is artificial, transitivity also plays an important role in implying extended connections among the Marvel character. It refers to how the relationship between two nodes in a network connected by an edge is transitive. By holistically analyzing the network and the most connected nodes, we hope to speculate plots among heroes/villains that can be developed in the Marvel cinematic universe.
Centrality is used as the fundamental theoretical framework to decipher the most popular characters in the Marvel Comics Universe. This information will help find the overlooked characters in the comics universe that could be introduced into the cinematic universe in the future as one of the business goals we would like to achieve in this paper.
Dataset Description
The dataset "Marvel character partnerships 2018" provided one JSON file that consisted of two parts, one is "nodes," and one is "links" category. According to the original description of the dataset, there are 350 nodes and 346 edges (or "links"). We divided this file into two different files. This action allows us to separate the nodes and edges into different files. Finally, we converted them into CSV file formation.
Nodes.csv contains a group, id, and size column. The group values include 0, 1, and 2. 0 is the group of heroes, 1 is the group of villains, and 2 is the group of antiheroes. A character is an antihero if they show qualities of both a hero and a villain. According to studiobinder.com, antiheroes are those characters who are notably lacking in heroic qualities. Sometimes they perform morally correct actions, often acting primarily out of self-interest or in ways that defy conventional ethical codes. The id column includes the names of the heroes, villains, and antiheroes. The size column tracks the number of connections that "id" or character has with others in the network.
For example, Baron Zemo is the first id and is categorized in group 1, which signifies he is a villain. Then we can see that he has two connections within the network.
Edges.csv contains a source and a target column. The file lists the connections of the IDs specified in the nodes.csv file. The source is the id from nodes.csv, while the target is the id of whom they are connected. The size amount from nodes.csv signifies the number of times an id will appear in the source column.
For example, Baron Zemo had a size of 2, so he will appear twice in the edges.csv with a different target value as shown above.
Analysis of the network
Analytics Tools
We used RStudio packages to explore, analyze and visualize the structure of the Marvel comic characters in the partnerships network.
Network Exploration
The Marvel Partnerships network is undirected with 350 nodes and 346 edges. It consists of 35% are heroes, 55% are villains, and 10% are antiheroes. The plot below (Figure 3) shows the distribution of characters in the Marvel Comic Partnerships Network. Based on observation, we can see most of the hero characters appear as groups or close to each other, and that concept applies the same to villains. However, antihero characters tend to stay in between heroes and villains or between villains instead of staying close together.
Five-number Summary
A small density result of 0.0056 also proves that many characters in the network are not connected or exchange information with each other.
Characters are connected in 56 different sub-communities (see Figure 3), and the majority of these groups do not have any connections between them. The largest community has 181 characters, and the smallest community has only two characters.
The tightly clustered nodes represent characters in one comic and do not have relationships or appearances in multiple comics. A moderate level of transitivity shows that are may occur some characters that appear in various comics and have cross-relationship or appearances in different comics and storylines.
Degree Distribution
The degree of a node in the Marvel Partnerships network is the connections it has to other nodes. The degree distribution is the probability distribution of these degrees over the whole network. In other words, the degree distribution represents the popularity and appearances of that characters in the network. With this in mind, we observe that the Marvel comic characters cross over with other characters.
The degree distribution in this network has a maximum degree of 12 and the minimum degree of 1. Below is the chart (Figure 4) ranking characters that have the most appearances in the network. Spider-man is the most cross-over with other characters, with 12 degrees; Captain America is the second most popular with 10 degrees; the third place is Red Skull with 9 degrees and so on.
The plot (figure 5) below shows the degree distribution of characters in the network with degrees greater than 3. We can see that the sub-communities on the right side do not have many characters with degrees greater than three compared to the large sub-communities on the left.
Subgroup level analysis - Community Detection
According to the book "A users guide to network analysis in R," modularity is an essential characteristic of a network used in many community detection algorithms. It is a measure of the network structure, specifically how nodes exhibit clustering where there is a greater density within the clusters and less density between them. We compare modularity scores and pick the algorithm that has the highest modularity
The result is shown in Table 2 below.
Comparing modularity scores between these algorithms, we can see that Louvain and Fast-greedy have the highest modularity scores. Though Louvain is slightly higher than Fast-greedy, we will use Louvain algorithms for our community detection. The plot below shows the community grouping using Louvain algorithms. We can see that the more prominent communities with some overlapping areas on the left side of the network. The Marvel Universe is constructed around the "us versus them" theme, suggesting close connections among heroes, with many villains as their common enemies.
Comparing modularity scores between these algorithms, we can see that Louvain and Fast-greedy have the highest modularity scores. Though Louvain is slightly higher than Fast-greedy, we will use Louvain algorithms for our community detection. The plot below shows the community grouping using Louvain algorithms. We can see that the more prominent communities with some overlapping areas on the left side of the network. The Marvel Universe is constructed around the "us versus them" theme, suggesting close connections among heroes, with many villains as their common enemies.
Subgroup level analysis – The largest sub-community
Based on the result of subgroup level analysis, we detected one community with the most significant number of community members that is community 4 (181 members). We separate this community from this network and continue to have a closer look at this group.
There are 64 heroes, 99 villains, and 18 antiheroes in this sub-community. Similar to the result in the whole network, we can see Spiderman, Captain American, Unicorn make the most appearances in the community.
Figure 7 is the visualization of the largest sub-community with 181 characters. The visualization is more precise now, showing the majority of heroes connected on one side and the other side are mostly villains. There are still some antiheroes in between some of the heroes' side and the villains' side. The further left to bottom left of the plot showing some appearances of one single or two heroes as they appear to work alone instead of as a group in the stories.
After separating this sub-community and run the 5-number summary and compare it to the whole network, we can see that the density is a little bit higher; however, it is still a small density score. The diameter and transitivity scores are the same compared to the whole network.
The degree distribution of this community is also similar to the whole network due to the highest appearances characters are in this community (figure 5). The bigger size of the node in figure 8, the higher the degree distribution.
We want to take a look at the betweenness of this sub-community to see what characters act as a bridge along the shortest path between two other characters.
As shown in figure 8 above, the characters have the most betweenness is Venom, second place is Doctor Doom, third place is Spider-Man, and so on. These characters tend to observe or control the flow of information within this community.
We also plot (figure 9) ranking characters with betweenness scores greater than 2000 and shown as the nodes' size (higher score = bigger node). The characters' names displayed are those characters that have a degree distribution greater than 5. This plot helps us see the distribution of the popular characters with the most control in the flow of information within this community. We can see that Spider-man is the bridge of the hero's group that controls the flow of information. Meanwhile, Venom is the bridge for information flow on the left side (villains' group).
We can also detect community or grouping within this sub-community.
The same concept with the whole network, we calculate and compare the modularity between the algorithms.
As a result, in table 4, the highest modularity belongs to the cluster_spinglass algorithm. We use this algorithm to plot and see how it detects the communities. Below is figure 10, the result of the community detection for this sub-community.
Future Discussion
The Marvel Partnership dataset offers several business insights on Marvel Cinematic Universe's future expansion based on the degree distribution. The following suggestions are all speculations based on our current understanding of the Marvel Cinematic Universe and our social network analysis results. We believe that these characters will play new or more important roles in the upcoming phase 4 Universe.
- Spider-Man: Spider-Man will continue becoming the bridge to connect existing plots and shape future plots of the avengers. Spiderman has been a long-standing cinematic figure in the past two decades. From the Spider-Man trilogy in 2002-2007 to The Amazing Spider-Man in 2012-2014, Marvel finally got the licensing from Sony in 2014. Spiderman has appeared in Captain America Civil War, two Avengers' movies, and formed two of its series, with an upcoming one in the fall of 2021 in addition to several animated spider-man series. The connections between Spiderman and other Marvel characters would offer different possibilities for Marvel to make storylines more intriguing.
Captain America: Although Captain America, Steve Rogers, was concluded in Avengers: Endgame, Marvel transitioned the storyline to Falcon, which led to appearances in the new - The Falcon and the Winter Soldier series in 2021. We believe Captain America will continue to play an important role in the phase 4 Universe.
- Unicorn: As a Soviet Intelligence agent assigned to security duty at the private laboratory of inventor Anton Vanko, Unicorn can appear as the new villain in the upcoming Black Widow, Shang-Chi, or future X-Men movies due to his relationships with characters in these films.
- Attuma: Attuma is an Atlantean nomadic warlord supervillain from Marvel Comics. With appearances from both Spider-Man and the ultimate villain for Avengers, we speculate that Attuma could play the final villain for the phase 4 Universe against the new Avengers.
- Venom and Loki: As the dataset was prepared in 2018, Venom and Loki have some of the highest degree distributions. Unsurprisingly, Venom and Loki will embrace their new productions in 2021. Venom 2 will be released on September 24, 2021, and a Loki series will be released on June 9, 2021.
The literature review referenced earlier explains that as a pure artificial network, the Marvel Character Partnerships network resembles a real social network. However, we learned that content creators must constantly consider the impact of characters' relationships on future plot progression. Therefore, the Marvel Character Partnerships network contains artificial manipulation to fit the storyline, and the construction of the character's relationship will not be entirely random.
Additionally, we found that even though villain and hero partnerships represent an intriguing part of Marvel Comic Universe, these relationships are rare cases. Loki and Venom are the most typical villains that have had convoluted relationships with antiheroes and heroes. Ultimately, heroes, as well as villains, tend to form their communities with their kind.
Conclusion
The Marvel Universe is one of the largest media franchises to exist globally. The relationships between heroes and villains have expanded over generations and have resulted in a network that crosses over Marvel Comics and Marvel Cinemas. The analysis of this network can answer the two important questions stated previously. One of the biggest uses for analyzing this network is to better the future comic/cinematic plans based on the previous connections and communities already associated with the network. For example, Figure 10 shows the sub-communities that can be seen within the largest community. There are roughly 16 sub-communities that can later be joined together into larger sub-communities by creating new partnerships between the characters in the network. Based on the size and contents of the network, the relationships and model can only tell us so much. We can make assumptions from the analysis of the network. Still, its biggest use will be to understand how the relationships cross storylines and how they will continue to expand the Marvel Universe and network.
References
Aslak, Ulf. "Marvel Character Partnerships (2018)." Https://Icon.colorado.edu/#!/Networks, 2018.
Heckmann, C., Chris Heckmann Chris Heckmann graduated from Emmanuel College with a Bachelor of Art, & Chris Heckmann graduated from Emmanuel College with a Bachelor of Arts degree in Writing. (2021, February 08). Anti heroes - 3 types that dominate film and tv. Retrieved May 11, 2021, from https://www.studiobinder.com/blog/what-is-an-anti-hero-definition/
Luke, D. A. (2015). A user's guide to network analysis in R. Springer.
Comments