Object Link
Description
Object Link enables the analysis of graph data. An example for such data is a bill of materials.
Classical Process Mining enables users to analyze single processes which consist of a linear sequence of activities. Examples of such processes are P2P, O2C or a single assembly line in a factory. The approach falls short if a complex network of processes and their interactions should be analyzed. Such process networks are wide spread as every organisation is not a set of independent sequential processes but actually consists of many processes which are interacting with each other. With Multi Event Log or Object Centric process mining, process networks and their interactions can be analyzed but they have to be modeled by hand. Each process step which is not a linear sequence, needs to be modeled into separate (case/object)-table. Therefore the complexity which can be modeled is limited. For example a bill of materials is by most source systems modeled as a recursively self referencing table, not as several (case/object) tables referencing each other. Object Link overcomes that limitation by reconstructing a graph out of the source data and providing functionality to analyze it.
In order to use Object Link, the source data to which is used to generate the process graph has to be marked in the data model editor. After the configuration is set the following functions are available:
LINK_FILTER filters along the links of the Object Link graph.
LINK_OBJECTS returns all objects of an input column that are part of the Object Link graph.
LINK_SOURCE / LINK_TARGET / LINK_ATTRIBUTES allows to access the source or target row from Object Link links.
LINK_PATH creates an activity table containing a case for each traversed path in the Object Link graph.
LINK_PATH_SOURCE / LINK_PATH_TARGET allows to access link attributes of LINK_PATH results.
The following paragraphs explain how you need to model your data in order to use the Object Link functionality. There are two different types of Object Link: Object Link and Signal Link. Since Object Link is intended to be a more generic approach, we will start with it and get to Signal Link at the end.
Configuring Object Link
Object Link should be used to model relationships between objects. There are two ways to create an Object Link configuration. The first way is to define the links between objects with so-called "signals". Signals are identifiers within rows that indicate which rows of tables are linked. The second way is to explicitly define the links between objects in a separate table called a mapping table.
To create the links using signals, we need to define one or more columns to identify outgoing signals and incoming signals of a table. These columns must be of the same data type. If two rows have a common signal identifier, i.e. one row has the identifier in the columns for outgoing signals and another row has the identifier in the columns for incoming signals, they will be connected by a link.
The following figure shows an example data model with two tables and three signals. Based on this configuration, a total of four links are created.
One data model with two tables and three signals. The arrows between the tables show the links in the graph.
The graph that was created based on the Object Link configuration.
The links can also be explicitly defined with a mapping table, where each row represents a link. The mapping table must contain columns with identifiers for the rows of the tables that we want to connect, at least one for the outgoing part of a link and one for the incoming part of a link. In addition, the tables that we want to connect must also contain a column with identifiers. The identifiers for each connected table must be unique.
The following figure shows how to create the previous graph using a mapping table:
The same Object Link configuration with a mapping table. Every row in the mapping table corresponds to one link in the graph.
As mentioned earlier, it is required that connected tables must not have the same identifier. To fulfill this condition, we can also specify more than one column that identifies the connection. For example, if column "Value" of "Table A" contains an identifier that is also present in column "Value" of "Table B", the tables cannot be connected. As a solution, a pair of columns with identifiers can be used.
The following figure shows the configuration with two mapping columns, which leads to the same graph as before:
The same Object Link configuration with multiple outgoing and incoming columns in the mapping table.
Object Link can be configured in the data model editor via the "Object Link" tab. If this tab is not visible, request "Object Link" from the service desk. Once Object Link is configured, with signals or with a mapping table, the data model must be reloaded for the changes to take effect.
Configuring Signal Link
Signal Link should be used to model relationships between activities instead of objects and unlike Object Link, only signals can be used to create the configuration, mapping tables are not supported.
Another important difference is that the LINK_PATH operator does currently not support Signal Link, only Object Link, and that the behavior of the LINK_FILTER operator is slightly different for Signal Link and Object Link. For more information, please refer to the respective documentation of the operators.
To create the Signal Link configuration with signals, we need to define the columns that identify the outgoing signals and incoming signals of a table, as in Object Link. The columns must be of the same data type, and if two activities have an identifier in common, they will be connected by a link.
The following figure shows an example where we create the same graph as before, but this time with Signal Link and on activity tables.
One data model with two activity tables and three signals. The arrows between the tables show the links in the graph.
Signal Link can be configured in the data model editor via the "Signal Link" tab. If this tab is not visible, request "Signal Link" from the service desk. Once Signal Link is configured, the data model must be reloaded for the changes to take effect.