 Welcome to this video introduction to network analysis with Gefi. This open source software is one of the best for network visualization. I find it easy to use. It leads us quite organically through the analysis process. This tutorial is based on two datasets to produce two different types of networks. You'll find these files as well as the complete process on the following page. I put the link in description. Note that this is simply a voiceover of my existing tutorial and the version with the audio is very often requested. Gefi has been updated several times since then but even if it brings important changes to the performance the interface is still the same and so this tutorial is still valid. But I'm planning new videos on Gefi in the near future so subscribe to this channel to stay informed. Before you start make sure you watch this video in high definition to see the details and you can turn on the subtitles to make sure you understand. So this is our first example, a one-mode network. Open Gefi. This is an older version. Now we have the 0.9.7. And before starting a new project we will import a plugin that we'll use for this example. This plugin is the GeoLayout plugin. You'll find it in the available plugin panel and search for GeoLayout. Here in the video I'm also importing the novel app because it's something that was not built in Gefi at this time but it is nowadays. So just GeoLayout and install. Accept the terms and then it will restart Gefi with this plugin. Restart now. Finish and reopen. Now you can create a new project and we'll use the first dataset. Go to the data laboratory and click on import spreadsheet. You will then navigate to your dataset, the nodes1.csv specify that we have semicolons between the values and import all the columns and specify the type of data you're importing there. Now we have this csv in our data table and we'll simply import also the edges table, which is edges1.csv and now we have this second file imported as well. Now all the action takes place on the overview panel. This is the place where the software will produce an overview of the graph specialized randomly and completely unreadable at first. Let's start by giving nodes a size proportional to their degree, the number of connections. In the ranking panel of the left column, select nodes and then this red diamond and then select degree in the rolling menu and enter the minimal and maximal value. I propose 10 to 100. Then let's specialize the graph. That's the main part. Let's begin with a specialization that gives more space to the graph in the graph in a decided area, the Früchte Mann Heingold, with the same values as in this model, 20,010. This visualization disposes nodes in a gravitational way. You're already able to distinguish community, more densely connected part of the network. Let the function run until the graph is stabilized, use the little blue magnifying glass at the bottom left of the graph panel to recenter the zoom. Then I propose to use the force at last two, another layout algorithm to disperse groups and give space around larger nodes. Be careful because the parameters you enter significantly alter the final appearance. My proposition here is to check prevent overlap and change the scaling to 50. Let the function run until the graph is mostly stabilized. Now let's have a look at the right side of this overview panel. You have this little statistics section. It will be very important to add some more information to the graph by giving the nodes new attributes so that we can influence the size, color of all these dots in the middle and also simply to explore the graph. In the statistics panel, click on average weighted degree to calculate this value for every node. The degree is not only calculated depending on simply the number of connection in this metric, but also what is incoming and upcoming and it's taking into account the value and the size or the weight of the edges. Now that these values are calculated, new attributes are available in the ranking panel. Select the color icon and choose weighted in degree to color nodes according to the number of incoming edges. I fine tune a bit the colors so that the scale will be more readable, especially because I know that few nodes have a highly weighted in degree. Now the result is that the biggest nodes, those with a high degree, are not necessarily with the biggest weighted in degree. Some nodes have lots of connections, but some of these connections could also be out connections, which means that now we have a color that gives us a different information from the size of the node. At the bottom right of the graph display, you'll find a little sign which allows you to develop a new panel. In Label, choose nodes to add the labels to your nodes and set their font, color and size. If needed, for example, if your data don't have any label column, click on configure to set the column content you want to get displayed. The ID may be used as a label, for example. And now we will try to finalize the graph. We'll go to preview for trimming the final details. Unlike during preview stages, changing settings in this menu is reversible and do not affect the structure of the graph. We're really just fine tuning the visualization itself. Here you find a few kind of suggestions of settings for good rendering. Like setting the edges opacity to 70% for a better contrast with the nodes. And you always need to hit refresh to get a new result, and it can take a few seconds to update after each change. At the bottom of this preview column, you'll find an export link. Note that exporting a PNG produces figure with a poor resolution. You may want to opt for an SVG, which is the advantage of being modifiable by your own image or drawing software. So we went through this analysis process extremely quickly, and of course, we want now to add some more information and to come back to the overview panel. Because, of course, visualization is only one step. Network analysis often needs other mathematical means to provide the researcher with a satisfactory result. Feel free to explore all the buttons of the statistics menu. For example, by playing with degree measures, density, path lengths, modularity. Here I propose to use this calculation called modularity, which is a way to highlight communities in the graph. So in this overview page, you just click on statistics, modularity, click on run to display this little window, and you choose a resolution between 0.1 and 2 and you click OK. And the next step takes place in the partition menu situated in the left column. You select nodes and modularity class in the rolling menu. You'll be then able to modify the colors attributed to the detected communities by clicking on them. Do not hesitate to repeat this operation with many resolutions. That's a research process. You want to be able to understand what modularity does. Another type of calculation could be the betweenness centrality, which measures all the shortest path between every pairs of nodes of the network and then counts how many times the node is on the shortest path between two others. It's a very interesting metric in the case of a network of letters, for example, letters that are sent and received because it allows the researcher to detect the people that occupy an intermediate position between two other people or groups. In the statistics panel, you can see that each node is on a network diameter and like the weighted indegree before, it's up to you to find a colorful way to highlight nodes that have a high betweenness centrality. It quickly appears that nodes with a high degree or weighted degree does not always have a high betweenness. During the import, you have noticed that every node was given a latitude and a longitude. The JLayout plugin will help you display nodes in a geographical way. In the layout panel, select JLayout and give it a scale of 20,000. Be sure that the plugin understands correctly that latitude is latitude and longitude is longitude. And set the projection to Mercator. You can adapt that, you can change, of course, but in this tutorial, I will provide you a Mercator background. As nodes are now grouped on a geographical coordinate, you'll have to give them some space. In the preview panel, check the final appearance of your network map and export it in SVG. You will then be able to import it on a background map. If you're familiar with Inkscape, download the map provided in the tutorial that has been created to fit with the chosen scale and Mercator projection. Open it and after having imported the map, you will be able to export it in SVG. And after having imported your network in it, select the city names layer and bring it to the front to make it readable. Feel free to try the same map with modularity. The result shows that communities are strongly related to geographic particularities. And then in the very end, do not forget to save your project exporting and saving the image file is, of course, a nice result, but of course, you need to save the Gefi file so that you will be able to come back to Gefi and change a few things and come back work on your project. Let's move on to our second dataset. The second dataset is a two-mode network, which means that it's a network of two different kind of entities. In the visual example here, you see the blue dots and the white dots. Imagine that it could be a network of, I don't know, companies and people and you will have people affiliated to companies. To deal with two-mode networks, you will need to install a new plugin. So open the plugin window and look for multi-mode network transformation. You will need to install it and restart Gefi. Now you can create a new project in the start window and go once again to the data laboratory, click on import spreadsheet to open the import window and import your first file. It's the file called nodes2.csv Specify that the separation between your column is expressed by a semicolon and do not forget to inform Gefi that the file you import is containing nodes, then press next and fill the import settings form has proposed. Inform Gefi that our cat variable is a string. This variable will be useful to separate members and companies in a further step. Follow the same procedure but with the edges file, edges2.csv and fill the form in the following manner, specify the semicolon and inform Gefi that you are importing the edges, fill in the last fields and uncheck create missing nodes because you've already imported them. Now go back to the overview panel. You're supposed to have 110 nodes and 142 edges. It's a much smaller network but of a very different nature. In the ranking panel give a size to your nodes here according to their degree between 10 and 50. In a tumor network, the degree centrality may not be a very interesting value because of the structural bias brought by the two different categories of nodes. In our case, the companies will be naturally much more connected than the members but this is just the first step. We just try to distinguish visually the two categories which is why we'll go now in the partition panel refresh the menu to make the nodes attribute appear and use the cat category to give a very different colors to the two type of nodes, members and institution and apply it to your network. Now deploy the network using the Forced Atlas 2 algorithm prevent node developing and scale it to 50. Your graph is now visually readable and looks very similar to many organizations network. It's a relatively small network so it looks quite good like that. You can stay like that if you want but it's also interesting to use this multi-mode network projection plugin to see how to go from a two-mode network to a one-mode network to project the graph. So you now use the multi-mode network projection panel available through the the plugin you downloaded and load attribute project the institution on the members if two members have an edge linking them with the same institution, they will now have a direct edge between them. Select the right attribute type, cat and set the matrix as proposed here member, institution and then institution member they must be symmetric with the type of node you want to keep at the beginning and the end. Check the remove edges and remove nodes buttons to clean the graph from the old institution's nodes and edges and finally click on run. Now the graph looks very different projection is always something that creates a lot of edges all the nodes that had connections to this institution are to the same institution now are connected together so it creates a lot of density. You can calculate the new degree centrality by clicking on average weighted degree in the statistics panel and then in the ranking panel you can apply this measure to the nodes then the new degree may be very different from the degree in the original network in the statistics panel click on network diameter to calculate between the centrality of your nodes then use this measure to color the nodes in such a network of people working in different committees or institutions or companies knowing who is at the intersection of two groups may be different for human resources officer etc. You can also change the color of the edges to make the multiple relations more visible in the final display I suggest here to give a strong black to all the edges that will be given one and then of course specialize the graph once again because it kept the positions of the nodes before the projection from a two-mode to one-mode network so you can use Fritiman Reingold or Fossilus2 and once again if you want to export this result you go to the preview panel you change the color of the edges so that it reflects the changes you made before and you export it Note that depending on your approach you may want to use this kind of network to analyze the neighbors of a specific node so you will first gray all the nodes and then use the little paint buckets on the left of the graph area and play with the tools on the top of this menu first paint the neighbors of neighbors of the nodes that is interesting to you and then the neighbors of this node in our example the red node member of only one committee is directly connected to 10 colleagues which are themselves connected to 49 other individuals and then once again you can export this result with the preview panel to save this new Gefi project as a Gefi file Thank you for your attention