Hi, and welcome to this lecture on social network analysis in ProM. So we’re still in the extension part of the process mining spectrum. And in this lecture, we will look at the social network part of a process. In ProM, there are different types of techniques that can help you with the social network analysis. And first I would like to explain the dedicated social network plugins. But we can also use a dotted chart with different settings, and even the inductive miner to analyze this aspect. But first, let’s explain the four different types of social networks that we can discover with dedicated plugins. The first one is hand-over of work.
So if we consider this event log on the top right, we recorded who executed it, we can analyze how work is handed over between the resources working in the process. So this is the formal definition. But in essence, it’s as simple as, for instance in this example, Anna handing over work to Bert. Because first Anna does something, followed by Bert. If you want to show this in a graph, we can do it as such. Each user is a circle, and we draw an arc whenever work is handed over between two users. For instance, Anna hands over work to Bert. And if we do this for the remainder of the trace, this is the full graph we get.
Bert hands over work to Dory. Bert hands over work to Chris, Chris to Bert, and Dory to Anna. And we can do this for all traces that we see in the event log. And we get a full graph knowing who hands over work to whom. The fourth type of social network actually requires specific recordings of when work has been reassigned. So for instance, activity a has the label reassign, meaning that Anna reassigned activity a to Bert. And this can indicate that Anna has a hierarchical relation with Bert. So in the graph, we denote this with an arc going from Anna to Bert. But actually, this information is rarely present in the event log.
And therefore, reassignment social networks are hard to discover. The third type is subcontract. Although it looks like reassignment, it’s a bit more subtle. So reassignment is actually when one user executes an activity in between another user. And In this example, for instance, Bert subcontracts work to Chris. So Bert subcontracts work to Chris, because Bert does an activity followed by Chris, followed by Bert again. And the fourth type is working together. Where, instead of handing over of work, we just record which users happen to work together on the same case. And in this example, all four users work together on this case. So we can draw arcs between all combinations of users.
But this also shows you which users, or groups of users, work on different types of cases compared to others. So knowing how these four types of social network analysis work, we can show this in ProM, as well as how you can use a dotted chart and inductive miner to cover the social network analysis aspect. So let’s switch to ProM and show you how it can be done. So let’s start with the BPI 2012 event log, only considering the a and the o activities. So let’s execute a plugins, and let’s search for social network. And now you will see the four types of plugins that we’ve discussed in the lecture. And let’s start with the hand-over of work.
Usually the default settings are OK. And when you do this on the real life event log, you get this. There’s not much to see. So here you can remove infrequent edges and see some colors up here. You can always play with the layout. Certain layouts work best for certain patterns that you can find. But here, for instance, the FR layout might help. And then you see a big cluster of users. Some are more on the edge, and these three are even further on the edge outside the usual resource cluster. Which means that they only from a few people get work assigned. And hence, have a typical, usually minimal role in the process.
And here are some options that you can use to change the visualization a little bit and to get different insights. Well, let’s see what else we can do on this event log using social network plugins. Well, as I said, the reassignment social network I can show you. But it will show you no edges, because no information has been recorded. So this means that no reassignment occurs between employees. Well, the subcontracting, you will get some network. And here you see that some users are not subcontracted to, but this user subcontracts to other users in the process. And actually, with some domain knowledge I know that this central user is a system. So the system does something, then another user does something.
And then the system is triggered again to perform an activity that has been recorded. So this nicely shows that certain activities are performed in a particular order.
The fourth type is the working together. And here we can see whether all users somehow or at one point in time work together on the same case. And even when removing edges, the colors stay the same. I can change the layout a little bit. But we see that it’s almost fully connected. So there’s an arrow between all nodes, which means that all users in one or more cases work together on the same case with another user. And as I said in the lecture, there are two more ways to analyze the social aspects of an event log. So let’s select the event log, visualize it, and then go to the dotted chart.
There are several ways how you can incorporate the social dimension in this. For instance, we can color it by the resource classifier. And then you see, for instance, that these activities are purple and these as well, which is the system user again. What you can also do is change the y-axis. Instead of the cases, you can change it to the resource. And every horizontal line has the same color. So every horizontal line is a particular resource. And if we now color by activity, we see that certain activities have been performed by certain users. And certain users have a preference to execute a particular activity. You also see that these users started later in the process.
Not much has been observed earlier. What’s more interesting is, if we change the time axis to, for instance, time since the week start. And then you’ll see that for instance the system works seven days a week. Most of the other users work six out of the seven days. And also if you choose a time dimension that since the day start, then you’ll see that most users start at 9 o’clock in the morning and work ends around 9 o’clock in the evening. And most users hopefully don’t spend 12 hours working. But you can see that this is not a typical business, that they
don’t stop working at 5:00 but they
continue working until 9:00. And by playing with the colors on the x and the y-axis, you can get some additional insights. Well, the third way is by using inductive visual miner. So let’s go back to the event log. And let’s start inductive visual miner. And let’s change the classifier to the resource classifier. Now you see the network from the resource perspective.
And you can see all resources are numbered. But you can see that these resources don’t perform work in a particular order. So instead of activities, we’re not trying to find order in the way activities are handled by the resources. And if we reduce the number of activities that we want to show, we see that there is some order. So user 112 is in the beginning, which is actually the system. But further, there’s no real process order to be observed between activities. And some event logs you might find something interesting here. That particular activities, or particular phases are usually executed by particular users. And all the usual visual inductive miner features are also available here.
So now you’ve seen how you can analyze the social network within the process using ProM. And there are three types of techniques; dedicated techniques that analyze how users collaborate; the dotted chart, which can give you insights in working patterns, and time and date patterns; and inductive miner, which shows you how cases are handed over between activities. So this was, again, an extension technique covering a different aspect of the process. In the remainder of this week, we will discuss other aspects of process mining, including example and case studies. Hope to see again in the next lectures.