Week 1 closing

Hi and welcome to the closing lecture of week one. In the beginning of week one, we’ve seen that event data is everywhere. Whenever you use your bank card, whenever you send or receive an email or telephone call, event data is recorded. Usually, nowadays also smart TVs record how they are used. As your smartphone, of course, also keeps track on how it’s used, and what apps are used, when and where. In more detail, we’ve discussed the public transport system, using public transport cards, and checking in and out, et cetera. This also creates a lot of event data that can be analyzed. And of course, visiting websites, even future learn. They also track a lot of event data that can be analyzed.

And all this can be used to improve the products. So, let’s look at a bit more detail at this public transport process and what events it creates. For instance, when you buy your public transport card, it can be recorded when you buy it, where you buy it, and what card number is attached to this. And also for instance how much you put on there. And then, whenever you check in or out, or whenever your top up, another event is created. That also contains where you did this and when. And using this we can reconstruct a process. So, in the public transport example, an event log consists of all the events related to a particular card.

And in this example, card 1337 contains several events. But also the card itself can have some properties. For instance, that it is prepaid. And then, each event contains some information. As we’ve seen before, the activity name, when and where it happened, et cetera, et cetera. And this together forms an event log. But also, we have other events and also other traces on this example card, for which we record events. And all this together forms an event log. So, what are the key ingredients for an event log? Well, we have a case notion and a case identifier. And the case can have a description and other attributes. So, in the public transport example, a public transport card was our case.

What happens to this card, where is it used, and when. And then each event contains the event name, or the activity that was executed, the timestamp when it was executed, but also the state. So, and event is an atomic observation. So, if I start and stop a particular activity, for instance buying the card could be recorded when I start buying it and when I stop, this would result in two events. The start and the completion of buying public transport card. Each having their own timestamp and maybe even resource. And, so in an event log, we have the case notion and the event notion. I’ve also shown you that this can be stored in the XES event log format.

And then, each case results in the trace of events. For a trace, we mainly have the ID, or the name, of the trace. And this is stored in the concept semicolon name attribute. For each event, we also have several attributes. Also the name or activity that is observed. And this is also stored in the concept semicolon name attribute. Then the timestamp when it occurred, but also the resource is recorded. That’s recording who executed the activity. And finally, in the life cycle semicolon transition attribute, it’s recorded what type of action was performed. So, was it the start or the completion transition of a particular activity. However, usually we don’t find data in ready-made event log format.

So, usually it’s in a tabular format, like this. And then we have to be able to detect which columns relate to which concepts in the event log. So, in this example, for instance, the case ID is the trace notion. Each row is an event and in the case column it’s stored for which card, or case, or trace, it’s recorded. And the second column, the action column, that contains actually the event name, which activity was executed. In the third column, the timestamp of the event is recorded. And these are the three key elements of an event log. This is the minimal requirement you need to do process mining. However, you can include more information.

So, for instance, in this table, also the location and the card balance is recorded. You can include this in your events and may be used for further analysis. I’ve also shown you that we will use ProM lite as our process mining tool. So, by now you have installed and use it for a couple of times, and then the remainder of this course we will show you which plugins you can use further to do process mining. ProM Lite, can be seen as a Swiss Army knife. It has some plug-ins, and in ProM Lite a particular selection of over 100 plug-ins is included, and each plugin provides a particular feature.

And together this forms a complete suite of process mining analysis that you can do. I’ve also shown you that ProM mainly evolves around three main views. You have the object view, where you can see whatever you have imported or created using plugins. The action view, where you can use objects to create other objects. And the visualization view, where you can look at the objects and inspect them. So, for instance, the dotted chart is a visualization of an event log object. And using these three views you can create new objects and perform a particular analysis. I’ve already mentioned it, we’ve also discussed a dotted chart, which shows a particular view on the event log.

And, this also explains very nicely several ways in which you can filter the event log, which we also have discussed. So, you can decide to keep or remove certain traces, but you can also do this on the event level. So, which events, based on activity included, or the time frame, which activities or events do you want to keep or remove. And this together, allows you to focus your analysis on a particular type of case, or a particular type of event. In this week, we’ve also seen the process mining framework and its positioning relative to the software system. So, the software system is the public transport card, or your phone or a website. And the world interacts with the software system.

You and me or other software systems. A process model is used to model how this software system should interact with the world. And while it’s interacting, events logs are created. And that’s the data that we use as input, and then we have to transform to event logs at some point. And process mining bridges the gap between this event data and the process models. So, using solely the event data, we can discover a process model. And we will discuss this in detail next week. Given a process model, either discovered or actually the one used to configure the system, we can do conformance checking. We can replay the data on top of this process model and indicate where deviations exist.

Finally, we can also extend a given process model. For instance, by projecting performance information on top of it, or by showing where deviations or conformance deviations exist.

During this course we will cover several process mining activities. So, we will start with the extraction phase, then discuss data preprocessing, which we already did in this week. And in the next week we will discuss several process mining techniques, discovery algorithms, for instance, the can help you in analyzing the process. We will also learn you the crucial part that’s evaluating the results. How does the discovered process model, for instance, relate to the observed data. And is the process model correct. And this loop has to be executed several times until you get results that you can summarize, and that can lead to process improvement. So, I hope to see you again in the next week.