Networks on – Who’s the Murderer?

Lab Reports, Networks, Visualization


Who’s the murderer is a Chinese variety online tv show produced by Mango TV since 2016. This show is originated from a Korean TV program that is called “Crime Scene”. In this show, celebrities from tv, film, and music industry will play detective and suspects in a game. Normally, one of the participants will be assigned a detective role. The others are all considered as suspects in the beginning but there is only one murderer. The goal for the detective as well as other participants is to find the real murderer. When the show started in season one, there is almost no connection between the characters in each episode. Each episode’s case is independent of the other. However, in season two, the show intentionally added more related scenarios and tried to build a universe for the whole program. For example, they create a crime story that happened in a fictional airline company in season one, and then they created another story that related to a competitor company in later seasons. Therefore, there is a possibility to build a networking map.


At the beginning of every show, the participant who plays the detective will draw the relationship map on a board so that they can develop the map as they recover new evidence hereafter. The post-production team will recreate the map with more detail and put it on their official website as well. The combination of these processes becomes the initial idea of my design.

an example of relationship map from S1E1

While I was looking for data set, I found another map that covers most of the network connections from season 1 to season 5. In this map, the blue and red lines indicate if the relationship between characters is a family or a lover, while the black line indicates other relationships. The map also uses shape and color to distinguish suspects, NPC, and detectives. To avoid spoiler, the designer of this map doesn’t show who is the murderer in the map. While this map is full of useful information and clear structure, I found that it also contains misplaced information. Meanwhile, since this map is not created by any software, it is hard for the author to maintain the data and design.

relation map

After I contacted the author of this map, I realized that she didn’t keep any data set. However, this map truly covers the data that I need. Therefore, I decided to create the data of my own.

Data Set

It took me a while to understand how to create a data set that can be used in Gephi. In this case, I created two separate spreadsheets. In the node sheet on the left, I create the ID, gender, and identity columns. This spreadsheet becomes the foundation of my network map. Then, in the edge sheet, I try to use the type to create a sense of direction. Unfortunately, I give up this function on a later design. Finally, I use 3 types of weight to describe the relationship. 1 = lover, 2 = family, and 3 = others.


Gephi has a well-organized working interface and I started with the data laboratory. It is simple to upload spreadsheet files separately and see the result. One thing that we should keep in mind is to add both files to the existing workspace. Otherwise, you can’t create a relationship between your file. After I uploaded the files, I found that I missed an edge. Unlike Tableau, Gephi provides an option for its user to modify data within the software. Therefore, I can manually add the edge.

Instruction on how to add an edge in Gephi

Next, we will move to the overview panel.

The most confusing design of Gephi is actually the overview panel. The user can’t see the actual result of their network map in this panel. Though the majority of the design is finished here. For example, we can add color to the node and the edge separately. In my case, I add color to the node to make a difference between gender.

Another function that I used is the statistic. However, to be honest, I think this function request the user to have more than basic knowledge of statistics. What’s more, the user needs to know his or her goal to better use this function. Therefore, though I tried this function and generated some reports, I can’t say that I understand it fully or handle this part.

The most important design process is to choose a layout for the network map from the overview panel. This part is also the biggest challenge for me. Most people will choose the Force Atlas layout at first to get an overview sense. However, while I was doing the test, the software crashed several times. And finally, we discovered that my data set is not suitable for this layout. Thus, I move to Fruchterman Reingold. layout.

initial layout
modified layout

As you can see, the first image is the initial result. Though it is not hard to read the connections and see the relationship weight, the map is a design disaster. The modified map is clean and still readable. However, the weight is not so prominent. Also, since the goal of this map is to show how characters from different seasons and episodes are connected, I want the design to show a cluster view instead of an equal distance. Therefore, I tried one more layout, which is the Yifan Hu layout.

This is the original result with color. I immediately noticed that one of the cluster is not connected with others. Therefore, I fixed the issues on data laboratory.

When I add the edge, the layout transfer to this shape. Though every node and edge is in the right place, I decide to manually change the spacing to make it more readable.

And here is the final map. I use less contrast color to change the feeling of the map. Then, the curve line add more sense of human connection to the map. The miss of legend here is quite frustrated. However, this map shows what I want.


Overall, this is a very changeling lab experience for me. I overcame lots of obstacles even from the dataset building stage. Then, since I can’t use the force atlas function, the process becomes much harder than I thought. As far as I recall, I restart the software at least 10 times during the design process. I also give up much information that I want to add to this map due to the limit of its function. For example, the color choice is not much and very hard to play with. I had to say that Gephi is not a new user-friendly software and anyone who wants to handle it needs much more time to learn to use it. Once you start to get familiar with this software, though, you will love it for what it can achieve.