Getting started

00:00: This video shows a first and very simple example of an object model. 00:04: It is intended as an introduction to the object analytics Paradigm. 00:08: You will learn how to define an object tree. 00:10: We will also show some first typical operations, namely, aggregations 00:15: along the edges of the object tree. 00:18: The data we use in this example, is from the BPI challenge, 2019. 00:22: It consists of a company's purchase orders where every purchase document consists 00:27: of a number of Articles or items. 00:30: Each of these items is tracked until it is received and runs through a series of process 00:34: activities performed by the company. 00:37: When building an object analytics model. The first thing to do is to Define the 00:41: root object, Think about what the core entity of your business is in 00:45: this example, it is natural to choose the purchase document as our 00:49: root object. 00:52: Each purchase document has one or multiple items, therefore it 00:56: makes sense to attach the items object to the root. 01:00: A purchase document has a number of properties that apply to the document as a whole 01:04: such as document type or spend area. 01:08: These Dimensions can be seen as features of the associated object. 01:13: Click on the spend area attribute to initiate a first query. 01:17: In the monitor window that opens you can see that in our example data. 01:20: There are 76349 purchase documents in total 01:24: with packaging being the most common spend area. 01:30: The item type is a natural property of the item if we open this attribute. 01:35: We see that there are 251734 items in 01:39: total. 01:41: Each item in turn has a number of events related to that item. 01:44: Naturally events is a sub-object or child object to items. 01:51: A property of an event is the activity type open the attribute. 01:56: In total we have 1.6 million events stored in our object model. 02:01: In this simple example, the object 3 is basically, just one branch starting 02:05: from the root purchase document to the leaf event. 02:09: You will soon see much more complex object models, but even in 02:13: this simple case the object analytics Paradigm offers some advantages. 02:19: Let's start with a first, very simple analysis, which illustrates the object 02:23: analytics paradigm. 02:26: Objects and sub objects are in a 1-2 n relationship. 02:28: Each document has several items and each item has several 02:33: events. Let's count the number of items for each document by aggregating 02:37: upwards along the hierarchies of the tree. 02:40: We can do this by dragging and dropping the items object upward onto the documents. 02:46: In the menu which appears we choose count as the aggregation function. 02:56: In the menu which appears we choose count as the aggregation function 03:00: and hit okay. 03:02: As a result we get a new dimension attached to the documents that counts the 03:06: number of items for each document. 03:09: Opening the default attribute 03:11: We see that most documents have only one item. 03:14: But there are also five documents that have more than 200 items. 03:19: We do the same for the events drag and drop the events onto the items to 03:23: count the number of events for each item. 03:37: The items object gets a new dimension the number of events of the item. 03:41: Open the attribute of the new dimension. 03:45: Typically an item has between two and five events. 03:49: We want to go one step further and instead of counting all events. 03:52: We want to count only events of a specific type. 03:57: We can do this by selecting the events, which should be counted for example record 04:01: Goods receipt. 04:04: Then start the aggregation Again by drag-and-drop. 04:15: We have gained a new dimension the name describes its semantics number 04:19: of events of type records Goods receipt of the item. 04:23: Open the attribute of that new dimension and remove the selection on the activity type. 04:28: You see that most items have only one Goods received event. 04:32: But there are also items with no receipt of goods. 04:34: How can it be that there is no Goods receipt. 04:38: Let's select these items by clicking on the bar. 04:41: And let's open the time of the items. 04:44: And switch the table to a chart. Many of the items have an 04:48: Associated time towards the end of the period. 04:51: It seems that the receipt of goods is still awaited. 04:54: However, some of the items are quite bold and still there is no receipt of goods. 04:59: How can this be? 05:01: Select these items for further investigation. 05:04: You will see that many of them also have an event called delete purchase order item. 05:09: Of course if the order was canceled, we would not expect to receive any 05:13: Goods. 05:15: But, what about the other? Very old items where the order has not been canceled, but there is 05:19: still no receipt of goods. Feel free to continue the analysis. 05:24: This has given you a basic idea of object. 05:26: Analytics what the object tree is and what entities are sub objects 05:30: are in the object model. You have seen a first typical operation in 05:35: an object model, an aggregation along, the hierarchies of the object tree. 05:40: As a first example in this video, we have seen how to investigate. 05:43: Why there are items with no Goods, receipts. 05:46: Note that we also have a more sophisticated version of this analysis ready for use. 05:52: Simply select load analysis. 05:55: And load our prepared analysis. 05:59: Click on continue in the top right corner. 06:02: This analysis allows you to choose items according to inclusion and exclusion criteria. 06:06: As in our example, where the exclusion was no receipt of goods. 06:11: Read the description fields to understand how to use this analysis, it 06:15: uses multiple selection sets. 06:17: You will learn more about selection sets soon. 06:21: This concludes the introduction to the business process data set in two 06:25: further introductory examples, you are going to see other very typical features of 06:29: object Analytics The relative time axis and sequence, analyzes.

00:00: In this video, we are going to introduce the relative time axis. 00:03: A very powerful tool in object Analytics. 00:06: It enables us to analyze different sub objects in relation to each other. 00:11: We will use it to compute time differences between certain events of Interest. 00:17: As described in the first video the root object in this example is the purchase 00:21: document with the recursive sub objects items and events. 00:25: The strength of object analytics lies in analyzing sub objects in relation 00:29: to each other. Here the object tree has only one branch. 00:33: But even in this simple case the object analytics Paradigm 00:37: offers some value namely when we analyze a subject in relation 00:41: to itself for example events and temporal relation to each other. 00:46: In this example, we analyze the time between the first and the last event. 00:52: Or the time between specific events, for example, the first event of type 00:56: create order item and the last event of type clear invoice. 01:00: We have already prepared an analysis for this purpose. 01:02: Let's load. 01:05: The analysis time between selectable events with rank. 01:09: Click on continue to move forward. 01:12: Everything is already prepared here in terms of the red selections choose 01:16: the starting event for example create purchase order item. 01:20: In terms of the yellow selections choose another type of events for 01:24: example clear invoice. The chart shows the distribution 01:28: of times between the first chosen start event to the chosen yellow end events. 01:33: There may be multiple of such and events for each item by means of the 01:37: dimension reverse rank of selected event. 01:39: You may in addition choose the last of the yellow events. 01:45: Play with the red and the yellow selections to choose different start and end events here 01:49: for example, invoice receipt to clear invoice. 01:55: Clean up. And let's see how we can get there from scratch. 01:59: The tool to do this is the relative time axis. 02:03: Open the activity type again. 02:06: We start creating the time axis by selecting the desired start event create 02:10: purchase order item. 02:13: Then click on the time icon at the time dimension in the events object. 02:18: And go to the detailed menu. 02:20: Default parameters are suggested here already the left hand side of the relative 02:24: time axis. The minuend is the event time. 02:29: Expand the menu 02:31: The right hand side, subtrahend is the time of the First event 02:35: of type create purchase order item. 02:39: The only choice which we need to correct is within which entity the time axis should 02:43: be computed. It should be the first event, within an item note 02:47: that there is a difference. If we sort all events within an item or within a purchase document, 02:52: And we put a meaningful name on the time, axis time distance to 02:56: Creation of order item. Then hit the okay button. 03:02: A new dimension pops up in the events object which assigns a time distance to 03:06: each event the distance to the first event of type create purchase order 03:10: item 03:12: In addition to the default attribute, we create a little nicer attribute to 03:16: show the time distribution in 10-day buckets. 03:19: Click to the settings icon. 03:22: Choose add attribute 03:24: We select a Time duration attribute and set appropriate parameters. 03:28: And we put a name on the new attribute before we hit the okay button. 03:33: Let's open the attribute on the canvas. 03:36: Switch to a chart 03:38: And choose the type of events for which we would like to see the time difference for 03:42: example clear invoice. 03:45: And here we are the time distribution between create purchase order 03:49: and clear invoice. Instead of clear invoice. 03:53: We can choose any sort of events for which to see the time distribution 03:57: towards that end of In addition we would now also 04:01: like to have a free choice of the start of event. 04:05: To get there. We create a second selection set. 04:08: Which we give the name start events. 04:11: Now we open the definition of the relative time axis. 04:15: And instead of a fixed semantics we switched to floating semantics. 04:19: The 0 point of the time axis should be whatever is currently selected in the red 04:23: selection set. 04:25: We adjust the name of the dimension accordingly and hit the okay button. 04:30: We open a second query window for the event type. 04:34: And assign the read selection set to this new query window by drag-and-drop. 04:54: Then choose for example vendor creates invoice. 04:58: Here we are choose any start event in the monitor to the left the 05:02: red selections. And any end event in the monitor to 05:06: the right the yellow selections? 05:09: With a little prettify at the analysis, shown at the very beginning. 05:13: The only thing we have not shown is how to build the rank dimension. 05:17: Look this up in the documentation video for the rank dimensions. 05:21: In theory, you could do the same analysis based on relational tables on 05:26: SQL. However, it will be very tedious to get there and 05:30: it will scale very badly in the amount of data. 05:33: Object analytics is designed to do that, kind of operations on massive amounts 05:37: of data. For example, 10 million patients with multiple billion of events 05:41: along lifetime. 05:44: You have now seen a second example of a typical operation on an object analytics 05:48: model the relative time axis. 05:51: It allows to analyze events streams in temporal relation to each other or 05:55: an event stream in relation to itself. Very often sub 05:59: objects to the root object are different sorts of events. 06:02: That means there is a time stamp in one of the next examples 06:06: you will see a sequence analysis another sort of analysis which 06:10: relies on time information.

00:00: This guide provides a walkthrough on analyzing sequences within an object analytics 00:04: model. In this example, we use business process data, specifically 00:09: the flow of events in processing. A purchase document. 00:13: As described in the first video, the root object in this example is the 00:17: purchase document. 00:20: With its sub objects items. 00:23: Each item in turn has events. 00:25: We want to understand what happens step by step with each item. 00:29: The flow of events or actions for each item. 00:33: Let's open the activity type. 00:38: Sort by frequency. 00:40: And select the most frequent events. 00:43: With the flow icon in the header. We can start the sequence analysis. 00:48: Because we want to see the flow of events associated with each item. 00:51: We need to set the items here as the embracing entity. 00:54: Furthermore, we want to see the first event before each event and 00:58: the five events after. This screen needs adjustment. 01:03: Alternatively, you can create the sequence via script, simply load 01:07: our script Repository 01:10: And open the prepared analysis. 01:13: Here is the sequence displayed in a tabular way? 01:16: Let's dive into a first analysis. For example, we might want to 01:20: analyze the context in which a delete purchase order item event appears. 01:24: Select that type of events. 01:27: The column minus 1 event in the table, show, the events 01:31: immediately prior to the delete purchase order item. 01:34: The column zero contains the events that are identical in time. 01:38: Usually, this is only the selected event, but in some cases, there 01:43: may be other simultaneous events. 01:46: Therefore, we chose delete purchase, order item also in column 0. 01:51: The further columns to the right show the events after deletion of an item. 01:56: You see that the event prior to the delete typically is create purchase order item. 02:00: That means often the item is created, but immediately deleted again. 02:05: But there are also cases where the prior event is receive order confirmation. 02:11: The first event after deletion is usually null, which means there 02:15: is no further event. This is probably what we would expect if the item 02:19: was deleted. 02:21: However, there are items where the next event is, vendor creates invoice. 02:25: Isn't that strange? The item is deleted but there is an invoice. 02:31: Select those cases to see what happens next. 02:34: And switch to the Sankey diagram. The Sankey diagram is sometimes 02:38: helpful to explore the process flow. Click the chart icon in the menu, 02:42: at the top to switch the table to the Sankey diagram. 02:50: Move the mouse over a specific event or transition in the diagram to highlight 02:54: what happens prior and after. 02:57: Typically after vendor creates invoice many items, go to record 03:01: Goods receipt 03:05: Then to record invoice receipt. Then the process works. 03:10: Some immediately go to clear invoice. 03:14: Some first to remove payment block. 03:17: But from their most of them also end up in clear invoice. 03:22: The item is deleted but it looks like the company is finally paying for it. 03:27: That's strange. Let's look at the supplier ID for these cases. 03:32: Most of the cases relate to vendor 0104. 03:35: We are not experts on this data, but it seems that the company should better 03:39: talk to the supplier. Explain data 03:43: is not a dedicated business process, mining software, but the generic 03:47: means of the object analytics Paradigm, make it possible to analyze this type 03:51: of data surprisingly easy. And you can easily add more data to the picture, 03:55: whatever additional data you have just hook them in as additional 03:59: subjects and easily extend the analysis. 04:03: In this video we have seen how to perform sequence analyzes in the X-Plane 04:07: data object Explorer. Note that event flows and sequences not 04:11: only occur in business process data, but virtually everywhere. 04:15: In particular in the medicine domain one is interested in the medication 04:19: sequence of patients who have series of diagnosis and prescriptions appended 04:24: to them. We are going to cover this example, in one of our next videos.