Iteractive PMCube Explorer Documetatio ad User Maual Thomas Vogelgesag Carl vo Ossietzky Uiversität Oldeburg December 9, 206
Cotets Itroductio 3 2 Applicatio Overview 4 3 Data Preparatio 6 3. Data Warehouse Schema........................... 6 3.2 Creatig a Meta-data File.......................... 9 3.2. Case Dimesios........................... 0 3.2.2 Evet Dimesios........................... 0 3.2.3 FactTable............................... 3.2.4 CaseTable............................... 2 3.2.5 EvetTable.............................. 3 3.2.6 Sematics............................... 4 3.3 Geerate Meta Data File From Data Source................ 4 3.4 Geerate Database Schema From Meta Data File............. 5 3.5 Import Data From Data Source to Stagig Area.............. 6 3.6 Itegrate Data From Stagig Area Ito Data Cube............ 6 3.7 Create Idexes o Data Cube Tables.................... 7 4 Applicatio settigs 8 4. Geeral Settigs................................ 8 4.2 Paths...................................... 8 4.3 Visualizatio................................. 8 4.4 Process Model Diff.............................. 9 4.5 Process Discovery............................... 9 4.6 Plugis..................................... 9 5 Coect to Database 20 2
Itroductio The Iteractive PMCube Explorer is a tool for multidimesioal process miig (MPM). This maual gives a brief itroductio of its usage ad the ecessary preparatio of evet data. For a geeral itroductio of MPM ad the cocept of the PMCube approach, we refer to [VA6]. For this documet, we assume that the reader is familiar with process miig. 3
2 Applicatio Overview Figure 2. shows the applicatio s mai widow after start-up. I geeral it cosists of four parts: the ribbo meu at the top, the cotet area at the ceter, ad two sidebars left-had ad right-had of the cotet area. Figure 2.: Mai widow after start-up The ribbo meu provides access to most of the features of the applicatio which are grouped i differet tabs by their sematic. The six mai tabs are: Start This tab offers geeral features like coectig the applicatio to the database (see Chapter 5) ad customizig the applicatio settigs (see Chapter 4). Tools This tab provides tools that support the itegratio of data ito the data cube like the editor for meta-data files (see Sectio 3.2). View O this tab, you ca chage the view of the applicatio, e.g., tur o/off the process model matrix view or hide/show particular tabs. OLAP This tab provides all OLAP-related features like loadig the cube ad its data from the data warehouse. 4
2 Applicatio Overview Evet Log O this tab, you fid all features to maipulate (e.g. filter) the evet logs that were loaded from the data warehouse. Process Miig This tab offers advaced process miig techiques like process ehacemet ad coformace checkig. Cosolidatio This tab provides access to the cosolidatio of process models. Tip: You ca hide/collapse the ribbo bar by clickig the ico o the right side (below the widow close butto) or double-clickig oe of the tabs. The sidebar o the left side of the mai widow comprises all cofiguratio optios for the OLAP query. Grouped by differet tabs, you ca roll-up/drill-dow or slice the data cube. Furthermore, you ca also omit uecessary attributes from the query results. I the sidebar o the right side of the mai widow, you fid the selectio ad cofiguratio of the process discovery algorithm, the selectio of the classifier, ad the operatio history where you ca udo/redo the aalysis steps. The cotet area i the ceter of the widow uses a tab system to preset the aalysis results. The Olap Cell Preview tab shows a preview of the curretly set OLAP query while the Olap Results tab presets a overview of the results of the last executed OLAP query. Tip: The mai widow is based o a dockig framework. Therefore, you ca customize it accordig to your eeds, e.g., by chagig the size of or rearragig the differet parts of the widow. 5
3 Data Preparatio The Iteractive PMCube Explorermaages the evet data required for process miig i a multidimesioal data cube that is stored i a relatioal data warehouse. Curretly, the followig database maagemet systems (DBMS) are supported: Microsoft SQL Server (recommeded) Oracle DB PostgreSQL MySQL However, it is recommeded to use Microsoft SQL Server which was primarily used for testig ad applicatio so far. Before aalyzig processes with Iteractive PMCube Explorer, the evet data has to be itegrated ito the data warehouse. Sectio 3. explais its expected database schema. 3. Data Warehouse Schema I Iteractive PMCube Explorer, there are two differet ways to model the case ad evet attributes. Dimesio A dimesio represets a axis of the data cube that ca be used to partitio the evet data ito sublogs. Each dimesio cosist of a set of pair-wise distict values. Furthermore, Iteractive PMCube Explorerallows the user to defie dimesio hierarchies i a tree-like structure. Dimesios should be the preferred way of modelig the attributes. However, they may ot be appropriate i case of may rare attribute values as ca this result i sparse data cubes ad missig represetativeess of of the process models. To avoid such problems, sets of differet attribute values may be grouped ito artificial categories, e.g. represetig itervals of values. If this is ot meaigful, the attributes ca also be stored as simple attributes. Simple attribute This way of modelig directly attaches a attribute to its case or evet. I cotrast to dimesios, simple attributes are very limited i their usage (e.g., they caot be used for roll-up ad drill-dow). Therefore, it is oly recommeded to use simple attributes for avoidig the loss of data if the attribute caot be meaigfully mapped to a dimesio. 6
3 Data Preparatio Colum Name(s) Type Cardiality Descriptio id arbitrary id of cell, must be uique dim... dim s arbitrary... foreig key to table of most fiegraied dimesio level Table 3.: Colums of Fact table The database schema maages case ad evet attributes o differet levels. This is ot oly reflected i the defiitio of OLAP operatios, but also by the database schema that stores cases ad evets i differet tables. Figure 3. shows the geeric database schema of the data warehouse which is a extesio of the traditioal sowflake schema commoly used for relatioal data warehouses. D.K m D s.k D s+.k x D w.k y D.K 0 D s.k 0 D s+.k 0 D w.k 0 id Fact Case Evet id id A A p A A r Figure 3.: Etity-relatioship model of geeric database schema Similar to the sowflake schema, there is oe table for each level of a dimesio hierarchy ad the tables of a dimesio are coected by a : foreig key relatioship accordig to the hierarchy (e.g., tables D.K 0,..., D.K m ). The Fact table refereces case dimesio (i.e., each dimesio represetig a case attribute) by a foreig key referece to the table of its most fie-graied dimesio level. Additioally, it also stores a uique id for the easy idetificatio of cell of the data cube. Table 3. gives a overview of the colums of the Fact table. Usig a foreig key, the cell id of the Fact table is refereced by the Case table which stores each sigle case as a row. Each tuple requires a madatory uique id to idetify 7
3 Data Preparatio Colum Name(s) Type Cardiality Descriptio id arbitrary id of case, must be uique cell id arbitrary foreig key referece to id colum of Fact table A... A p arbitrary 0... colums for storig simple case attributes Table 3.2: Colums of Case table Colum Name(s) Type Cardiality Descriptio id arbitrary id of evet, must be uique case id arbitrary foreig key referece to id colum of Case table dim... dim w arbitrary... foreig key to table of most fiegraied dimesio level A... A p arbitrary 0... colums for storig simple evet attributes Table 3.3: Colums of Evet table the cases. Additioally, the Case table ca also have a arbitrary umber of additioal attributes (A,..., A p ) to store the simple case attributes (i.e., all case attributes that are modeled as simple attributes). The simple attribute colums of the Case table may have arbitrary types. Table 3.2 gives a overview of the colums of the Case table. The Evet table stores oe evet per row. Similar to the Case table, it refereces the tables of the most fie-graied dimesio levels of all evet dimesios (i.e., evet attributes that are modeled as dimesios). The simple evet attributes (i.e., evet attributes modeled as simple attributes) are represeted by a additioal table colum for each simple attribute (similar to Case table). The Evet table also has a uique id ad a foreig key referece to the id of the case that ow the evet. Table 3.3 summarizes the colums of the Evet table. The structure of a dimesio level table depeds o whether the dimesio level is the highest (i.e., most coarse-graied) level of the dimesio. If it is ot the highest dimesio level, the table has a uique id colum (used for referecig), a value colum (holdig a represetatio of the attribute s value), a descriptio colum (providig a more verbose textual descriptio of the value), ad a paret id colum holdig a foreig key referece to the id of the paret dimesio level value. The latter ca be omitted for the highest level of a dimesio, because this does ot have a paret. Table 3.4 gives a overview of the colums of a dimesio level table. Iteractive PMCube Explorerprovides a set of tools that support the creatio of the data cube ad the itegratio of data. With these tools, you ca create a meta data file from scratch (Sectio 3.2) geerate a meta data file from a selected source file (XES oly) (Sectio 3.3) 8
3 Data Preparatio Colum Name(s) Type Cardiality Descriptio id arbitrary id of dimesio value, must be uique value arbitrary represetatio of dimesio level value descriptio varchar/varchar2 textual descriptio of the dimesio level vale paret id arbitrary 0... foreig key referece to the table of paret dimesio level, ca be omitted if o paret exists Table 3.4: Colums of Evet table geerate a database schema for the data cube from a selected meta data file (Sectio 3.4) import the data from data source ito a temporary stagig area table of the database (Sectio 3.5) trasform the data of the stagig area table ito the cube structure (Sectio 3.6) create idexes o the data cube to improve database look-up performace (sectio 3.7) You ca fid the data itegratio tools o the Tools tab of the applicatio s ribbo meu (see Figure 3.2). Figure 3.2: Tools tab Adapters for accessig differet data source are provided as plug-is. Curretly, there are adapters for XES ad CSV files available. The followig sectios explai how to use these tools. 3.2 Creatig a Meta-data File To be able to load the evet data from the database, Iteractive PMCube Explorerrequires some meta-data o the database schema which is provided by a XML-based meta-data file. To create this file, Iteractive PMCube Explorer provides a meta-data editor that ca be opeed from the Tools tab (see Figure 3.2). The editor widow provides a umber of differet tabs to edit the meta-data file which are explaied i the followig i more detail. 9
3.2. Case Dimesios 3 Data Preparatio O the Case Dimesios tab (see Figure 3.3), all dimesio levels ad their relatioships eed to be defied. Dimesios ca be added by typig the dimesio s ame ito the text iput field o the left bottom ad clickig o add dimesio. The dimesio ad its logical root (ALL) are added to the tree-view o the left. After selectig the dimesio ode, oe ca chage its ame ad its data type (default is strig). Figure 3.3: Meta-data editor (Case dimesios) Selectig the ALL ode ad a right-clickig o it opes a pop-up meu that allows to add a ew dimesio level as a child of the selected ode. After selectig a dimesio level i the tree, oe eed to defie the properties as specified i Table 3.5. Note: To create a dimesio hierarchy, the dimesio levels must be specified from top to bottom (startig with the most coarse-graied dimesio level). 3.2.2 Evet Dimesios O the Evet Dimesios tab, oe eeds to specify the evet dimesios ad their dimesio levels. The workflow ad the properties are similar to the Case Dimesios tab (see Sectio 3.2.). 0
3 Data Preparatio Property Dimesio level ame Dimesio level table ame Id colum ame Value colum ame Descriptio colum ame Foreig key to paret id colum ame What to eter Name of the dimesio level (as it should be displayed i the tool) Name of the database table represetig the dimesio level Name of the colum holdig the id Name of the colum holdig the value Name of the colum holdig the descriptio Name of the colum holdig the foreig key to the paret table (ca be left empty if o paret table exists) 3.2.3 FactTable Table 3.5: Properties of dimesio level O the FactTable tab, oe has to specify the table ame of the Fact table ad the ame of the colum that holds the cell id (cf. Figure 3.4). Furthermore, oe eeds to specify the foreig key relatioships of the Fact table to the tables represetig the dimesio levels. Clickig the Add butto opes a ew dialog to create this mappig (cf. Figure 3.5). Figure 3.4: Meta-data editor (FactTable) I the first field, oe eeds to eter the ame of the colum holdig the foreig key to the dimesio level table. For the secod filed, click o the right-had butto ad
3 Data Preparatio select the dimesio level that is refereced by the foreig key from the dialog. Cofirm the mappig by clickig OK. Figure 3.5: Meta-data editor (attribute editor) 3.2.4 CaseTable O the CaseTable tab (Figure 3.6), you eed to specify the structure of the Case table by eterig the followig iformatio:. The table ame of the Case table Figure 3.6: Meta-data editor (Case table) 2. The ame of the colum that holds the case id (primary key) 3. the ame of the colum that holds the cell id (foreig key to Fact table) 4. For each simple case attribute stored i the Case table, eter the colum ame ad data type of the attribute ad press Add 2
3.2.5 EvetTable 3 Data Preparatio O the EvetTable tab (Figure 3.7), you must specify the structure of the Evet table. Eter the followig iformatio: Figure 3.7: Meta-data editor (Evet table). The table ame of the Evet table 2. The ame of the colum that holds the evet id (primary key) 3. The ame of the colum that holds the case id (foreig key to Evet table) 4. For each simple evet attribute stored i the Evet table, eter the colum ame ad data type of the attribute ad press Add 5. For each evet dimesio, create a mappig betwee a dimesio ad the correspodig foreig key colum i the Evet table 6. Select the attribute that should be used to order the evets (usually the timestamp) 7. Select the attribute that should be used as the default classifier (usually the activity) 8. For each simple evet attribute stored i the Evet table, specify the aggregatio fuctio that should be used for evet aggregatio 3
3.2.6 Sematics 3 Data Preparatio O the Sematics tab (Figure 3.8), you must specify the sematics of particular attributes (similar to a cocept i XES) especially for the time-stamp. Click Add ad eter the followig iformatio i the dialog:. The attribute that should have the sematic 2. The cocept the selected attribute should have For the time-stamp, eter a format strig required for parsig the time-stamp which describes the time-stamp format of the attribute i the database. The format strig must be defied accordig to the C# time format specificatio (https://msd.microsoft. com/de-de/library/8kb3ddd4(v=vs.0).aspx) Figure 3.8: Meta-data editor (Sematics) 3.3 Geerate Meta Data File From Data Source Istead of creatig a meta data file by had, you ca also derive it from the structure of the data source file by performig the followig steps.. Start the wizard by selectig Geerate meta data from the Tools tab. 4
3 Data Preparatio 2. Select the adapter matchig the type of your data source (e.g., XES adapter) ad click Next. 3. Select the data source file from your disk. Optioally, you eed to adjust the adapter optios. Click Next. 4. Select the attribute which represets your case id from the drop dow meu ad click Next. 5. Select the orderig attribute (typically the time-stamp attribute) ad the default classifier (the attribute represetig the ame of your evet). The, click Next. 6. Select the file system locatio ad the file ame for the ew meta data file. 7. Click Fiish to create ad save the file. 3.4 Geerate Database Schema From Meta Data File Iteractive PMCube Explorer provides a tool to automatically create a database schema from a existig meta data file. This file ca either be extracted from the data source (see Sectio 3.3) or maually created (see Sectio 3.2). Note: The tool assumes a already existig ad empty database. You ca prepare a ew database usig the admiistratio tools of your database maagemet system. Start the wizard by selectig Schema Geerator from the Tools tab. Select the meta data file from your file system. Select the already cofigured coectio to your database or create a ew oe (see 5 for details). Click OK to create the schema i your database. Tip: Alteratively, you ca also write the scripts for creatig the database schema to a file ad maually execute it later. However, you have to select the type of your target database as the geerated scripts may cotai vedorspecific SQL code. 5
3 Data Preparatio 3.5 Import Data From Data Source to Stagig Area To load the data from a data source ito the database, Iteractive PMCube Explorer provides a import wizard. Usig the specific adapter plug-is, it reads the data from file ad stores it i a flatteed database table (called stagig area table). For the trasformatio of the data ito the cube structure, aother tool is provided (see Sectio 3.6). To load the, do the followig:. Start the wizard by selectig Import data to stagig area from the Tools tab. 2. Select the adapter matchig the type of your data source (e.g., XES adapter) ad click Next. 3. Select the coectio to your database ad click Next. 4. Select the data source file ad optioally adjust the adapter s optios. The, click Next. 5. Cofigure the data import: a) Eter a ame for the stagig area table (accordig to the table ame guidelies of the used database maagemet system). b) Select all attributes of your data source that you wat to load ito the database (by default, all attributes are selected). c) For each selected attribute, cofigure the respective colum ame i the stagig area ad its data type (optioal, i most cases, default values should be okay). 6. Click Next ad wait util the import is completed. 7. Click Fiish to close the wizard. 3.6 Itegrate Data From Stagig Area Ito Data Cube After you have loaded the data ito the stagig area table, you ca trasform it ito the multidimesioal structure of the data cube. Iteractive PMCube Explorer supports this with a tool allowig you to map the colums of the stagig area table to the dimesios ad simple attributes of the cube. To trasform the data, do the followig:. Start the wizard by selectig Itegrate data ito data cube from the Tools tab. 2. Select the meta data file from the file system ad click Next. 3. Select the coectio to your database ad click Next. 4. Fid the stagig area table i the table list (left-had) ad expad it. 6
3 Data Preparatio 5. For each colum of the stagig area table, map it to its correspodet data cube elemet: Select the colum ame from the list. Select the target elemet by clickig o the related ico o the right-side (i.e., case dimesio, simple case attribute, case id, evet dimesio, simple evet attribute). 6. If all colums are mapped to their target cube elemet, click Next to proceed. 7. Deselect a particular data itegratio step if you wat to skip it (typically all steps should be performed). 8. Click Next ad wait util the processig is completed. 9. Click Fiish to close the wizard. 3.7 Create Idexes o Data Cube Tables To improve the performace of the data look-up, you ca optioally idex the data cube s tables. Therefore, you have to perform the followig steps:. Start the wizard by selectig Idex data cube from the Tools tab. 2. Select the meta data file from the file system ad click Next. 3. Select the coectio to your database ad click Fiish. 7
4 Applicatio settigs Before startig the first aalysis, it is recommeded to cofigure the tool. To customize the applicatio settigs, click o the Applicatio settigs butto i the start tab. The applicatio settigs dialog showig up provides multiple settigs o differet pages. The followig sectios will give a brief overview of the available cofiguratio optios. Note: Chages of some particular settigs will oly take effect after a restart of the applicatio. 4. Geeral Settigs O this page, you ca select the cofiguratio file for the logger. Iteractive PMCube Exploreruses the Apache log4et library (a.net port of the log4j library). To eable the loggig, create a cofiguratio file (see log4et documetatio for details) ad select it i the applicatio settigs. Additioally, you ca also set maximum tree depth of XML documets hadled by the serializer. However, the default value should be fie i most cases. 4.2 Paths O the Paths page, you ca cofigure the default paths for the import ad export files (e.g., process models, images), a default folder for storig meta-data files, ad the default folder for the settig files of process miig algorithms. Furthermore, you have to specify the file that will store your database coectio iformatio. 4.3 Visualizatio O this page, you ca customize the appearace of the process model visualizatio (e.g., ode size, distaces betwee odes etc.). Most of the settigs are specific for a particular process model type. To edit the appearace of a specific view model, you ca select the correspodig process model otatio from the drop-dow list. https://loggig.apache.org/log4et/ 8
4.4 Process Model Diff 4 Applicatio settigs O this page, you ca select the algorithm that should be used to create differece models. 4.5 Process Discovery Here, you ca select the default process discovery algorithm. This algorithm will be automatically selected o applicatio start-up. Optioally, you ca also select a cofiguratio file to provide customized default settigs to the selected default algorithm. 4.6 Plugis The Plugis page gives you a overview of all plug-is that have bee loaded at the start-up of the applicatio. To search for a particular plug-i, the text field o the top of the page allows you to filter the plug-i list. Tip: By default, the tool provides multiple plug-is. Therefore, a empty plugi list idicates that the path to the plug-i folder is ot correctly set. To fix this issue, you ca select the plug-i directory ad cofirm it by clickig the Okay butto. Now, the plug-is should be loaded whe restartig the applicatio. 9
5 Coect to Database At the begiig of your aalysis, you eed to select the coectio to your database. To ope the coectio dialog (Figure 5.), click o the Coect butto of the Start tab. Figure 5.: Coectio Dialog To create a ew database coectio, select the used database maagemet system from the drop-dow list ad click o the plus ico. After that, eter the required parameters i the dialog. Table 3.2. explais the parameters i more detail. For coectios to Microsoft SQL Server databases, there is also the optio to use the automatic autheticatio of the Widows operatig system. You ca check the checkbox to eable this optio. User ad password field ca be left empty whe Widows autheticatio is used. You ca save the settigs usig the Save ico i the toolbar o top of the dialog. 20
5 Coect to Database Property Coectio ame Database Server Port User Password Descriptio The user-defied ame of the coectio which is used i the coectio list o the left. Should be uique. The ame of the target database The URL or IP address of the target database server The port of the target database service The ame of the database user The password of the database user Table 5.: Parameters of the coectio dialog Note: The database cofiguratios are saved i a file (for selectig this file, see Sectio 4.2) which is ecrypted usig the Widows user accout. Cosequetly, the coectio file ca oly be read from the user accout you created it. Sharig coectio iformatio betwee differet user accouts or machies is ot possible. Fially, you eed to select the meta-data file that describes the structure of the selected database. To coect to the database, select the desired coectio from the list ad click o OK. Now the applicatio iitializes the dimesio values from the database. Depedig o the size of the data ad your coectio speed, this may take some time. Tip: The applicatio automatically saves the last used database coectio ad meta-data file. To recoect quickly, you ca click o the Quick coect butto i the Start tab. 2
Bibliography [VA6] Thomas Vogelgesag ad Has-Jürge Appelrath. PMCube: A Data- Warehouse-Based Approach for Multidimesioal Process Miig, pages 67 78. Spriger Iteratioal Publishig, Cham, 206. 22