Tolkien User Guide
Contents
1. Tolkien User Guide June 2001 perotsystems 8 Table of Contents 1 USING TOLKIEN eC 1 INTRODUCTION REPRHNERS 1 1 2 HANDEING PROJECTS M OR a RR RENE de ac was 2 12 1 Starting a new project iss ese ti eee tee dete thee hie dre gebe tede desee a scent dg 2 1 2 0 Opening an existing project 3 1 2 3 Deleting an existing project iiie oeste esta is EEE E E oen a 3 1 3 HAND NG DATA MODELS 5 corretti rte nee Her EH ER eek x Er dot es 3 13 1 Creating a new data model aiit m eet ete d ent 3 13 2 Copying a dataanodels 3 italienne e tara bare 3 13 3 Deleting a data modella idt ee teli anten 3 1 4 HANDLING TABLES RELATIONS AND 4 1 4 1 Drawing and editing tables 4 1 42 Drawing and editing relations ss 5 1 5 CONNECTING TO A DATABASE TO IMPORT A DATABASE SCHEMA sr 6 1 6 CREATING CDB Li CODE ie tercie eerte tere dy eee norte fetes 7 1 7 PROPOSITIONALIZATION OF A DATA MODEL uses serrer enne enne 8 1 7 1 Viewing preprocessing trees anona auratis BH PEG HERO 9 1 7 2 Executing the propositionalization operator ss 11 2 QUESTIONS AND REMARKS 5 2 eerie e rette osos ees so oo 12 3 ABOUT SYLLOGIC INNOVATIONS eeeeeeeeee essen stessa tnnt tn s2. Connecting to a database to import a database schema In stead of drawing the database schema you can also import it from an ORACLE database other types of databases are not supported at this time To do this click the Database connect button Figure 5 The Database connect button You will be asked to provide a name for a new Database model which will be created in the model tree and which will be used to import the data into After providing a model name the Connect dialog will appear Date 06 14 01 Page 6 Document version 1 1 perotsystems rs nnect Database name Imydatabase Host name ravenstein Port number 1521 User name USERO Password eMe Ok Cmm Figure 6 The Connect dialog The Connect dialog appears with a Database name Host name Port number User name and Password field The standard port number to connect to an Oracle database is 1521 The input you provide is case sensitive Keep this in mind when for example entering the user name When you have entered correct values see the file tnsnames ora in your Oracle installation for the hostname and port number of a database all tables from the database will be displayed in a list You can select the tables you want to import use the SHIFT and or CTRL keys to select or deselect one or more tables All relations for which you import the from and to table will also
3. Server When tables attributes or relations are created on the client they are directly stored in this server Therefore there is no need for a Save option to save a project Date 06 14 01 Page 2 Document version 1 1 perotsystems 1 2 2 Opening an existing project Select the Open Project menu option from the File menu A list of existing projects will be shown with their names and their creation dates Select a project and press OK The data will be read from the J2EE Server this may take some time and the tree of models will be shown Select a model in the tree to view the corresponding model on the right side of the screen 1 2 3 Deleting an existing project In order to be able to delete an entire project that project must be the current project Select the project name in the root of the tree of models and press DELETE or select the DELETE menu item from the Edit menu 1 3 Data models 1 3 1 Creating a new data model When a project is available new data models can be created in the tree of models Select the base for the new data model in the tree of models The base of a data model can be the root model or a previously created data model Now select the Add new Model menu item from the Tools menu or right click on the base position in the tree and select the New X model menu item from the popup menu the X here stands for the type of model that can be added You will be ask
4. an attribute by clicking on it the Edit and Delete buttons become active With the Edit button you can change the name and or type of the attribute with the Delete button you can remove an attribute from the table Check the Target table checkbox if you want this table to be the target and select a target attribute from the Target Attribute list The Non data list can be confusing at first It is a list of attributes of the table Only selected attributes in this list are considered to be non data attributes for example primary or foreign keys are non data attributes In this way the data or non data property of the attributes of a table can easily be set Use Shift click to select more adjacent attributes Use CTRL click to select or deselect non adjacent attributes You must provide a name for the tables before you can generate CDBL code A table must also have attributes in order to be able to set it as a target table 1 4 2 Drawing and editing relations To add a relation between two tables hold down the control key e click on one of the tables and press the mouse button e draga line to the other table while holding the control key To add a relation to a table itself right click on the table and select the Add relation to itself option from the popup menu Date 06 14 01 Page 5 Document version 1 1 perotsystems To edit the properties of a relation do
5. be imported Standard all attributes from a table are shown on the screen These attributes can be hidden as described in the previous section 1 6 Creating CDBL code CDBL stands for Common Declarative Bias Language It can be used to describe a data model It can be used by ILP engines for data mining When you want to generate CDBL code for a certain model first select that model from the tree of models If a model is not empty CDBL code can be generated for it rp BL Figure 7 The Generate CDBL code button The CDBL code appears in a Generated CDBL Code window You can save the generated code to a file by clicking the Export button The button lets you go back to the worksheet Date 06 14 01 Page 7 Document version 1 1 perotsystems E enerated DBI code CDBL code for amp data mining project OR_Test1 b data model Import CDBL version 1 1 Generated on Mon Jan 08 13 12 45 CET 2001 b Generated by Syllogic s Mining Mart Model Server gt E 6 Begin of generated CDEL code rget MM PROJECT STARTDATE ableMM PROJECT l PROJECTID PROJECTNAME DESCRIF ON STARTDATE ENDDATE CREATIONDATE ROOTMODE p lable MM MODEL l HODELID MODELNAME DESCRIPTION REATIONDATE PROJECT lypesiMM PROJECT numeric nominal nominal nominal nomin ial nominal numeric Eme Figure 8 The Generated CDBL code window You can a
6. can be drawn between two tables and associations can also be drawn to a table itself Furthermore tables and associations can be repositioned To draw an association hold down the CTRL key while having the mouse above a table Press the mouse button and drag the association to the other table From every model so called CDBL code Common Declarative Bias Language can be generated This is code that describes the model It could be used for multi relational data mining using an ILP algorithm The University of Leuven Belgium for example has an ILP algorithm that can read this code Others are expected to follow Further a propositionalization operator is provided This can be used to propositionalize parts of a database schema after which a mining application can be used to mine the created result table The remaining part of this chapter focuses in more detail on the different parts of the program 1 2 Projects 1 2 1 Starting a new project To start a new project choose New Project on the File menu A project name will be asked and after pressing this name will be shown as the root of a tree of models at the left side of the screen the right side of the screen will be empty At this point new data models can be created or a database model can be imported from an Oracle database A project consists of data models which in turn consist of tables and relations objects are stored in a Java 2 Enterprise Edition
7. N_DATE GENDER a rae o A Target Region ACCOUNTID LOAN DATE OWNERTYPE etape LOAN_DURATION LiveNpADistrict LOAN MP DISTRICT NAME REGION CITIES AVERAGE SALARY RATIO URB INH LOAN OK DISPONENT Figure 1 Screen shot of the Tolkien client application 1 WARNING Tolkien is a demonstrator program and not intended to be used in a production environment It is still in development and although the functionality that is described in this document should work be aware that bugs may occur when using this program Date 06 14 01 Page 1 Document version 1 1 perotsystems Tolkien can be used to draw different types of data models e Database Model shows how the model is stored in a database e Conceptual Model this is a cleaned up version of the database model Things in the model that are not relevant to the concept but were for example added for data management reasons can be left out here e Business Model this is a part of the conceptual model that focuses on a certain business area e Analysis Model this is a part of the business model that will be focussed on for further analysis These models are displayed in a tree which makes it easy to keep an overview of the created models One can start by importing the database model from an existing database Oracle or by drawing a model yourself Tables attributes and associations can be added and edited More than one association
8. be dragged to a different position when they are selected H Figure 2 The add table icon By right clicking on a table you get a pop up menu with a few options Target attribute Edit By selecting Edit or by double clicking a table you can change the properties of that table Hide Show attributes By selecting this option the attributes of the table are shown or not shown on the screen depending on the status Set target Set Target lets you select an attribute of the table as target attribute Add relation to itself By selecting this option a relation from the table to itself can be created Delete Delete deletes the table Table name Attributes Target table Non data Figure 3 The Edit Table dialog Date 06 14 01 Page 4 Document version 1 1 perotsystems The Edit Table dialog With this dialog you can change things like the table name and the attributes of the table In the Table name field you can set a name for the table Note that actions like changing the name of a table or adding or changing attributes only change the stored data model these actions have no direct effect on the tables or attributes that are stored in a database The Attributes list contains all attributes of the table You can manipulate this list with the buttons at the bottom of the list The New button lets you add an attribute simply type the name and choose a type When you select
9. ed to provide a name for the new model and after pressing the model will be added to the tree 1 3 2 Copying a data model When working with a tree of data models it is very useful to be able to copy a data model to a sub data model or another data model somewhere in the tree of data models One may for example import a database schema into a database model copy that model to a conceptual model and then edit it until it reflects the conceptual model To copy a data model right click it and choose Copy to from the popup menu A dialog will appear and a target data model can be chosen The copied tables and relations will not be related in any way to the original tables and relations 1 3 3 Deleting a data model To delete a data model and its submodels from the tree of models select the data model in the tree and press DELETE or select the DELETE menu item from the Edit menu Be sure that no part of the model is selected on the right side If a table or relation was selected these will be deleted instead Date 06 14 01 Page 3 Document version 1 1 perotsystems 1 4 Tables relations and attributes 1 4 1 Drawing and editing tables The add table icon adds a new table to the worksheet A table is represented by a UML class diagram figure One or more tables can be selected by clicking on a table or by dragging a selection box around one or more tables the selected tables have a red border Tables can
10. ement that describes the action The preprocessing tree will not be executed on the database at this point This is an action that is left to the user and is described further on 1 7 1 Viewing preprocessing trees At some point of time in a project a data model may or may not have preprocessing trees associated with it This can be viewed by right clicking a data model and selecting List available preprocessing trees If no preprocessing tree is associated with the model a message will be shown that there are no trees If only one preprocessing tree is available then that one will be displayed directly If more preprocessing trees are available the user is presented a list to choose from Tree description Nr 1 Auto generated Tree with target table MOC Nr 2 Auto generated Tree with target table MOC Nr 3 Auto generated Tree with target table MOC Figure 10 Dialog for choosing a preprocessing tree After choosing a tree this tree will be shown Date 06 14 01 Page 9 Document version 1 1 Figure 11 Example of a preprocessing tree for a model An example of a preprocessing tree is shown in figure 11 The example shows a tree consisting of two preprocessing actions The properties of an action can be viewed by double clicking on it An example is shown in figure 12 Action Primary input Secondaryinput Output Tree Action id 545 Action type 5 Definition SELECT MODEL tMODEL ID MODEL_
11. etate tn seins 13 perotsystems 1 Using Tolkien 1 1 Introduction In a data mining task a lot of time is spend on the preprocessing of data This is often a quite technical task that requires knowledge of various fields It is a process that should become easier faster less technical and with more focus on the business domain that is involved An environment that uses a visual language to capture domain knowledge could help to make it easier to apply domain knowledge in the data mining process Furthermore operators could be developed that can automate part of the preprocessing process The Tolkien program is a demonstrator program that takes a small step in this direction Figure 1 shows a screen shot of the application The program provides the following functionality e import function for importing part of a database schema e creation and manipulation of data models e propositionalization operator for automatic propositionalization of multi relational data e generation and export of CDBL code cUm Piel E3 File Edit Tools Help OBZ 9 perotsystems S ei Rs Project MyProject Model type Database Model Model name Databese Mode 9 7 aConceptualModel aBusinessModel CLIENT anAnalysisModel DISTRICTID ACCOUNTID E CustomerData cuento 1 5 ACCOUNTID 9 Conceptual SEX FREQUENCY pan ARIS creditcardUsage AGE CREATIO
12. ith the name of the output table Date 06 14 01 Page 11 Document version 1 1 perotsystems 2 Questions and remarks If you have any questions or remarks about this demonstration application please send a message to marc dehaas ps net or olaf rem ps net Date 06 14 01 Page 12 Document version 1 1 perotsystems 3 About Syllogic Innovations Syllogic Innovations is the innovative department of Perot Systems Nederland Syllogic Innovations has expertise in among other things system integration database management data warehousing machine learning data mining and text mining This expertise is gathered in products such as Adaptive Systems Management and solutions like Prognostics and Health Management Perot Systems is a world wide supplier of information technology and business solutions The corporation provides its clients with an integrated service offering a combination of multiple disciplines that assist our clients in improving their business operations or in creating new business offerings Perot Systems delivers these integrated services through creative partnerships that are broad and unique fostering opportunities for innovation new levels of profit and growth and the creation of new types of business Date 06 14 01 Page 13 Document version 1 1
13. lso copy text from the CDBL Display window and then paste it into another application 1 7 Propositionalization of a data model Tolkien provides a propositionalization operator for handling multi relational data This operator uses aggregates to capture information from various tables and merge this information into a single output table After selecting a target table and a target attribute one can propositionalize a data model by right clicking the data model and choosing Propositionalize data model The following dialog will appear Figure 9 Dialog for entering propositionalization information Date 06 14 01 Page 8 Document version 1 1 perotsystems This dialog has a Depth Database name Host name Port number User name and Password field The depth parameter tells the operator which tables to include in the propositionalization process A depth value of one means the operator will only include tables that have a direct connection to the target table A depth value of two means that also tables that are connected via another table to the target table are included in the process The meaning of the other fields will be clear Remember that the input is case sensitive The operator will generate a preprocessing tree A preprocessing tree is a tree of preprocessing actions The actions have a primary and secondary input table a definition and an output table The definition is a SQL stat
14. tIS_MUTAGEN MODEL tLUMO MODEL tLOGP count cnt ATOM r max ATOM_r0 CHARGE max CHARGE ATOM r0 min ATOM r0 CHARGE min CHARGE ATOM r Sum ATOM r0 CHARGE sum CHARGE ATOM r avg ATOM r CHARGE avg CHARGE ATOM ATOM r2 ELEMENT pred tpred ELEMENT ATOM r2 pred ELEMENT ATOM r2 ATOM r2 TYPE pred tpred TYPE ATOM r2 pred TYPE ATOM r2 MODEL MODE t ATOM ATAM N Remark The action definition can be edited and copied but changes will not be Ok Figure 12 Dialog showing properties for a preprocessing action Date 06 14 01 Page 10 Document version 1 1 perotsystems 1 7 2 Executing the propositionalization operator After propositionalizing a data model a preprocessing tree is generated The SQL code that is defined in the actions of the tree is not executed automatically on the database containing the data model This can be done by clicking the Execute SQL from tree button or selecting the Execute SQL menu item from the Tools menu SaL b Figure 13 The execute SQL button The button and the menu item are active only when a preprocessing tree is displayed After clicking the button or selecting the menu item the dialog of figure 9 is shown again The Depth parameter cannot be edited at this point but the user should provide the information for making the connection to the database When the resulting table has been created a message will be shown w
15. uble click it or right click the relation and select Edit from the popup menu An Edit Relation dialog appears here you can change the name of the relation and other properties Relation name SYS C002943 Table MM MODEL Table MM PROJECT Attribute PROJECT v Attribute PROJECTID v Multiplicity Om v Multiplicity 1 v Direction Both v OK Cancel Figure 4 The Edit Relation dialog The Relation name field contains the name of the relation by default NewRelation Use this field to change the relation name to something more meaningful The two Table fields contain the from and to table of the relation these can never be edited The attributes that are linked by the relation can be altered by selecting an attribute from the Attribute drop down list The corresponding multiplicities can be changed in the same way The direction in which the attribute can be used can also be changed from Both to Left to right or Right to left Left to right means from the left hand side attribute to the right hand side attribute in the Edit Relation dialog Left to right and Right to left do not correspond to positions of tables on the worksheet You must provide a name for your relations before you can generate CDBL code Both tables must have at least one attribute in order to be able to draw a relation between them 1 5
Download Pdf Manuals
Related Search