AU2007216642A1 - Data Mining System - Google Patents

Description

2007 16:47 Wal ington-Dummer Sydney No 1634 P. 3 P 30/009 Regulation 3.2 0 0 ci o 0

AUSTRALIA

Patents Act 1990 COMPLETE PATENT SPECIFICATION Invention Title: DATA MINING SYSTEM The invention is described in the following statement including the best mel of performing it known to the applicant: Our Ref: 072028 COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:47 Wallington-Dummer Sydney No.1634 P. 4 2 DATA MINING SYSTEM 0 0- The present invention relates to computer based ci systems for data manipulation and, more particularly, to

C)

processes sometimes known as data mining.

C< BACKGROUND

VO

\O Computers commonly store large amounts of data which Ci contain inherent and latent relational patterns and which o are potentially valuable in providing basis for managerial ci and operational decision making. Yet such data is often widely distributed within and amongst databases so that the extraction and use of such patterns and relationships is not readily available. Hence numerous data mining systems have been developed which seek to interrogate relevant databases using criteria to bring such relational patterns to light. However Data mining (DM) may be defined as discovering profile or behaviour patterns of customers, clients and other entities to better understand and subsequently better serve them in a more efficient or profitable manner. Transactions from databases are mined i.e. amalgamated, sifted, probed and analysed using specialist software. The ideal outcome is a set of business targets, such as a list of customers that you are currently at risk of losing and whom you must fight hard to maintain.

Data mining has had patchy success because it is hard to use, does not generate actionable results, and has poor COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:48 Wallinston-Dummer Sydney No.1634 P. 3cohesion to the business function that it is designed to C( serve. Data mining products use an archaic paradign that Sreflects their research heritage, and data mining as a

C)

VS discipline has not developed pragmatic methods to de rease project risk. Taking each of these in turn: j Hard to use: current technologies focus on

NO

VD manipulation of a few technical details to the p- exclusion of other practical aspects. Users struggle Sto manipulate the software, often not understanding which operations they need to perform or how the controls fit into those tasks. Users are often not confident that the results yielded from the software are true or accurate, and they are hesitant to defend those results when challenged.

Does not generate actionable results: this is biggest single cause of data mining project failure. A manager will present some vague business objectives; the analyst will then take those away and perform some technical operations, injecting their own assumptions and inconsistencies. By the time the analyst returns with a result it bears little resemblance to what the manager intended or contains items that the business cannot do anything practical about.

Poor cohesion to the business function that it is designed to serve: over time the work performed with COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:48 Wal ington-Dummer Sydney No-163C P. 6 4 Co 0\ the data mining software becomes brittle, irrelevant, or lost. People forget what they d, inherit some work left by a departing staff m and they have no idea how it works or what it 5 The results that it generates have little conn to the main business of the department or compan software, together with any preceding work, idle, unused items.

It is an object of at least some embodiments present invention to address or at least ameliorate s the above disadvantages.

Notes 1. The term "comprising" (and grammatical vari thereof) is used in this specification in the inc sense of "having" or "including", and not i exclusive sense of "consisting only of".

2. The above discussion of the prior art in the Back of the invention, is not an admission thainformation discussed therein is citable prior z part of the common general knowledge of persons s in the art in any country.

BRIEF DESCRIPTION OF INVENTION Terminology: in this specification an Activity diagi also known as a Transform diagram. A Relationship d is also known as a Match diagram.

stale, or amber, does.

action j. The Decome >f the Dme of ations Lusive n the ground any .rt or <illed am is iagram COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:48 Wai inston-Dummer Sydney No.1634 P. 7 o Accordingly, in one broad form of the invention there CA is provided a method for data mining of at least one i database by means of computer-implemented software; said VS method including the steps of: creating at least one task defining Document for each of said at least one task,

VO

O defining within said Document a Business Rules p diagram for said at least one task, S(c) defining within said Document at least one Technical Operations diagram for implementation of Business Rules of said Business Rules diagram defining a Source Data icon indicating location of said at least one database or data file, executing said Technical Operations with said Source Data to generate at least one output diagram, verify that said at least one output complies with said Business Rules; and wherein each said diagram is a graphical rendition constructed by means of said computer-implemented software; each said diagram convertible to executable code adapted for processing by a central processor of a computer system.

Preferably said method includes the further s:ep of defining within said Document data for a Test Rig diagram to satisfy said Business Rules.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05

I

2007 16:48 Wallington-Dummer Sydney No.1634 P. 8 6 Preferably said method includes the further step of

C

verifying correct functionality by application of said at Sleast one Technical Operations diagram to said data of said n Test Rig diagram, and wherein each said diagram is a graphical rendition constructed by means of said computerimplemented software; each said diagram convertible to

VO

IN executable code adapted for processing by a central processor of a computer system.

0 o Preferably said Document is composed by means of a user interface display generated on a display device linked to said computer and wherein descriptive and annctative text sections may be defined with said document.

Preferably said interface display comprises at least, a Document construction region, a Resource library region and a common productivity accessory region.

Preferably said Document construction region is adapted to accept a combination of text and "drag and drop" Resources accessed from said Resource library area.

Preferably one or more Resources are combined into a diagram in said Document construction region; each said diagram representing a subtask.

Preferably at least one said diagram is a Business Rules defining diagram.

Preferably at least one said diagram is a Technical Operations diagram.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:48 Wali ngton-Dummer Sydney No.1634 P. 9 7 o Preferably said Technical Operations diagram may comprise an activity diagram, a relationship diagram or a Scombination of activity and relationship diagrams.

lr Preferably a technical operation diagram may link in other technical operations diagrams which will embed and execute together when the former is run.

\O

IN Preferably said Test Rig diagram comprises a sample of input data and a sample of output data; said input data and o said output data adapted to verification of one of said Business Rules and/or validation of one of said Technical Operations diagrams.

In a further broad form of the invention there is provided a computer-based data mining system wherein data mining is performed according to at least one user-defined rule for at least one associated data mining task; said system including a rule testing process wherein a sample of input data and a sample of expected output data are adapted to said at least one rule; said at least one rule implemented through a Document based diagram structure wherein each of at least one diagram of said diagram structure is translated into a computational process by said system.

Preferably said user-defined rule is a formulation of a characteristic of interest sought in Source Data for a data mining operation.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 20D7 16:49 b.Se. 207 1:49 Walngton-Dumr Sydney o64P~ No.1634 P. 8oPreferably said system includes construction of CATechnical operations diagrams; said diagrams including relationship and activity diagrams.

Preferably said relationship diagrams represent a userdefined relationship between sets of Source Data.

Preferably said activity diagrams represent user- Va defined processes applicable to said sets of Source Data.

Preferably each of said diagrams is constructed by a o user in a Document; said Document provided as zuser interface on a computer display.

Preferably said document is a readily interpretedi corporate record of the business and technical steps involved that may be discussed, annotated, archived, reviewed, revised within the business operations.

Preferably each said diagram is translated by software of said data mining system into executable code for processing.

Preferably said user interface includes Libraries of Resources; said Resources including data mining operations and application activities.

Preferably said user interf ace includes produ ctivity accessories; said accessories including calculator, a database diagnostic tool and statistical functions.

BRIEF DESCRIPTION OF DRAWINGS OOMS ID No: ARcS-i 59837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:49 Wallington-Dummer Sydney No-163K P. 11 9 1^ 0 0 ci

C)

Co Embodiments of the present invention will n described with reference to the accompanying dr wherein: Figure 1A is a representation of a computer syst implementation of the data mining system of the p invention, Figure 1B is a flowchart of the basic ste implementation of a preferred embodiment of a data operation according to the present invention, Figure 2 is a view of a user interface screen dis on a personal computer of the computer system of Figuz Figure 3 is an example of a document constructed user interface of Figure 2, Figure 4 shows a list of diagrams associated wit of five data mining processes, Figure 5 shows a table for use in defining a rules for performing a data mining operation, Figure 6 shows a relationship diagram implementation by the software of the data mining sysl Figure 7 shows an activity diagram for implemer by the software of the data mining system, Figure 8 shows an example of a library of Resourc use in construction of the relationship and ac diagrams of Figures 6 and 7, Ow be awings m for resent Os of nining played e 1, in the h each for -em, tation es for tivity COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:49 Wal ington-Dummer Sydney No.1634 P. 12 io 0 Figure 9 is an example of a series of business rules 0C for a data mining project entered into the table of Figure

C)

V Figure 10 is a set of input data for use with a test rig, Figure 11 is a set of expected output data resulting

O

O from the operation of the test rig, Figure 12 is an example of an overarching project o diagram for coordinating a number of subtasks in the data mining operation, Figure 13 is an example of a portion of a result table generated by the software of the data mining system according to the invention, Figure 14 is an example of an activity diagram in accordance with a preferred embodiment of the data mining system of the present invention.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:49 Wallington-Dummer Sydney No-1634 P. 13 11 o DETAILED DESCRIPTION OF PREFERRED EMBODIMENTS ci i In broad terms embodiments of the present invention

CD)

comprise a document centred data analysis software system.

Users develop data mining solutions by drafting eC conventional business documents containing text and tables MO that describe the business situation. They embed active

NO

Cq data mining content containing queries that are run against Stheir database to produce actual results for those situations. The document integrates business-focussed discussion and executable technical operations. The document is also a common language that allows analysts and managers to clearly communicate with each other aboit the task they are performing.

With reference to Figure 1, a computer implemented system 10 for mining data from a variety of computer stored databases 12, includes at least one personal computer 14 interfaced with a server 16. The system includes a scftware application in which a user (not shown) is presented with a user interface 20, shown in Figure 2, which permies the construction of sophisticated criteria and procedures for interrogating multiple data sources stored on the system.

Logically, a given data mining operation is divided into a number of subtasks. Each subtask may be defined in the following steps, to be explained in more detail b low: define a Business Rules diagram for the sub ask, COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:49 Wal inston-Dummer Sydney No.1634 P. 14 12 o Optionally define a Test Rig diagram to satisfy C( the Business Rules. This involves selecting input and expected output values,

C)

S(c) define Technical Operations diagram(sl to implement Business Rules. These are either transform or match diagrams,

VO

VD Optionally connect the Test Rig to the Technical SOperations and run tests to verify correct o functionality, define Source Data diagram the location of actual data to be mined (which may be the output of other subtasks). These are either transform or match diagrams, run Technical Operations with Source Data to generate Results diagram(s), Verify that Results meet Business Rules.

Results generated by a given subtask may be used as input for the Business Rules of other subtasks. Transform and match diagrams may be reused as either operati>ns or source data of other sub tasks.

Each defined subtask is identified by a subtask name 22 associated with a Document 24 which defines it. The name is displayed on a tab 26 of the user interface 20. A toolbar button (not shown) may be used to execute the Document constructed by the user by means of the user interface COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:50 Wal inston-Dummer Sydney N o -16 4 P 13 o The term "Document" in this description refers to a 0 conventional computer-based document which can includ text

C)

and drag and drop icons.

n With reference to Figure 2, a user is provided with tools to construct a Document within the displayed user Sinterface 20. The interface 20 is divided into three

VO

Sseparate areas A, B and C. Area is the working space in which the actual Document is constructed. Area contains 0 o Libraries 28 containing icons representing Resources which may be accessed and dragged onto the Document construction area Area is reserved for common productivity accessories including calculator, a database diagnostic tool and statistical functions. In an alternative preferred form the productivity accessories in Area may be incorporated in a tab in the libraries area This is a tab on the libraries area in which several accessories are available: Table calculator that performs operations whenever cells in a table on the document are selected. It computes the sum, minimum, maximum, count, standard deviation, range, etc of those selected cells.

Bookmarks list.

"To do" list.

List of currently running Diagrams with progress indicators and controls to cancel each one individually.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:50 Wal inston-Dummer Sydney No.163W P. 16 14 l o Resources 0 (C The Resources of area include data processing operations, data mining algorithms, data table and

C)

V~ external information (variables) and other functions organized in Libraries of Resources represented Sdiagrammatically by icons 28. Several Resources nay be VO functionally linked together in the Document area A, to form a solution for a subtask. A user drags a selection of o Resources from Libraries into a diagram 30 in the Document area A as shown in Figure 3. Diagrams may then be linked by a "click-and-drag" process to form a complex useful function.

Each Resource has specific settings that define its operation and which can be accessed by the user by means of pop-up windows under the Resource icon. Settings nay be displayed or hidden as desired.

Resources may also take the form of Templates which contain skeletal outlines of common business or other application situations. The user drags a selected Template into the Document area and fills in fields of the Template with his or her own data. Once on a Document, a Template can be edited to suit the user's particular requirements. A range of Templates may be provided with the data mining system to suit a variety of business and other data related applications.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:50 Wallington-Dummer Sydney No.1634 P. 17 15 o Elements O 1. Elements are building blocks. They have the following

CD)

features.

busine es database table

(N

0 a. Several elements are connected together on a Diagram to form a solution for a given small-scale proalem or granular piece of the task.

b. Elements is an umbrella term covering all runnable building blocks, including: i. Data processing operations ii. Data mining algorithms iii. Data tables from databases or text files iv. Parameters (variables) v. Reused Transform or Match Diagrams c. The user drags an Element from a Library onto a Diagram. Elements are then connected together by clicking-and-dragging between them. The Diagram combines several connected Elements into a complex, useful function.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:50 Wallington-Dummer Sydney No.1634 P. 18 16 0 (C d. Each Element has specific settings that define its Soperation and the user can access these settings as l V a pop up window under the icon. Settings can b displayed or hidden under each Element as desired by the user.

VO

VO

The Document Feature SThe Document, which is a central feature of the 0 (C 10 present system will now be described in greater detaLl. As noted, it is a conventional Document which is constructed by a user using text and combinations of the Resources available from the Libraries in area B of the user interface to diagrammatically represent a particular business or other problem.

As shown in Figure 3, each diagram 30 (representing an arbitrary number of such diagrams) is constructed in the Document in the form of a boxed field 32 and represents a particular executable task defining any of the five tasks denoted 44 in Fig- 4. Thus a diagram may contain a Technical Operation or a user-defined business rule or test rig or result. User-constructed diagrams which may be useful for future data mining operations within an organization, may be saved and added into a Defined Resource library. Diagrams of different types may be linked together within a Document to give more complex operations COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 200 1^ 0 0 ci

C)

Co 07 16:51 Wa linston-Dummer Sydney No.1634 17 and criteria for mining than can be achieved wit single diagram.

HTML text can be inserted into the areas 38 between and below the diagrams 30 in the Document as in Figure 3, to provide comments, explanation: contextual information. The text area 42 at the top Document may contain a summary of the overall subtask addressed by the Document.

More particularly: a. Each document integrates several aspects.

i. It is a conventional document which th< drafts and can be used as a bu communication tool. It uses text, tabl drag and drop icons to diagrammat represent the business problem. The docum a common language that allows analyst managers to clearly communicate with each about the task they are performing.

ii. These diagrams are executed within Agil Mining software as data mining code, pro actual results.

iii. The modular arrangement of diagrams an text areas within the document gives rise "executive summary working details" la, which business-centric diagrams and te placed at the top of the document P. 19 hin a Lnd shown and ,f the to be user siness !s and ically ent is s and other Data ducing i HTML to an out in ct are and COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:51 Wa lington-Dummer Sydney No.1634 P. -18 o operation-centric diagrams and text are placed C at the bottom.

Siv. Both linking executable diagrams and linking n text through HTML hyperlinks (see below) means that documents form a "wet" of intertwined business context, runnable

O

diagrams, and results, creating an integrated, multi-faceted solution to a complex business o problem.

b. The physical structure of the document is: i. A user interface that appears as a series of pages that can be scrolled from top to bottom.

The pages appear "joined" by dotted lines that mark the bottom of a previous page and the top of the subsequent page. The pages contain a vertical sequence of diagrams and HTM: text areas.

ii. Each diagram is indicated by a box border and constructed by dragging Elements (icons from libraries) onto the document constructicn area and connecting them together. Each diagram is a standalone artefact representing a specific Subtask within the overall project and may contain both technical operations and business rules.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:51 Wa l ington-Dummer Sydney No.1634 P. 21 19 o iii. These diagrams link with other diagrams to form C conglomerations of tasks that eventually define the entire project. Diagrams may be n additionally saved in the Custom Made Elements library. Diagrams of different types nay be Slinked to each other, giving a broader and more

VO

flexible view than could be achieved with one diagram type alone. That is, two individual Stasks can be concatenated by linking their diagrams. When executed the software will complete both tasks, with output data from the first task flowing as input data to the second task. There is no limit to the number of diagrams that can be linked in this way.

iv. HTML text areas can be drag-and-dropped between Diagrams and at the top of the Document. The user types comments, explanations, and context information in these areas. The text area at the top of the Document is for holding a summary of the task that the whole document addresses. Text areas can be added under each diagram as necessary to hold explanations and notes specifically for that diagram, as well as HTML hyperlink cross references to other related diagrams and Documents.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:51 Wallinston-Dummer Sydney No.1634 P. 22 20 o c. Each document contains a Task, which is a self- C' contained unit of work for the project. The:e are six considerations to be taken into account when o n solving a data mining problem. These are: i. Business Rules: the business criteria that the task must meet. This aspect ensures that

VO

the work is relevant and usable for improving business operations.

Sii. Test Rigs: show that the task meets both business requirements and is correctly implemented. The tests specify concrete examples to compliment business rules. They also verify correctness of technical operations by showing that the solution works as expected.

iii. Data transformation sequences: technical operations are arranged in a series of steps. For many data mining operations a sequence of calculations is the easiest way to work with the data.

iv. Data matching between tables: technical operations are arranged as a set of tables with relationship connectors between them, e.g. field A equals field B. Often information from complex data structures is most easily extracted from this view.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:51 Wallington-Dummer Sydney No.1634 P. 23 21 0 v. Results: output from technical operations is integrated into the document to show the n eventual outcome of the work. The user visually relates the results to the above ci Sconsiderations to ensure that they are met.

vi. Creating models from the data: this Ls the aspect of detecting higher-level behavioural characteristics using data mining modelling algorithms. Model building is seen as a distinct step compared to transforming and matching data it generates a new model element compared to manipulating existing data. This is often thought of as a factory operation: submit data and pick up model.

d. Individual diagrams are used to address each of these considerations; they are described in the following paragraph.

e. There are six distinct Agile Data Mining® diagrams.

i. Business Rules Diagram.

business titeris UntitledDiagrami .3 as.p COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:52 WalIington-Dummer Sydney No.1634 P. 24 22 S i. A table in which the user defines on rule 0 (N per row. A business rule is a c: ncise definition of a single specific bu iness situation and optionally how to handle it, 0 e.g. a high value customer is ona who ci purchases more than $1000 per year. Rules Va V.0 are entered either as text or dragged rom a ci previously defined rule (Templates library, 0 Custom Made Elements library, or other ci diagram).

ii. Initially used for planning and defining the scope of each document. This gives finegrain business instructions, enabling accurate development of each Transform, Match or Model Diagram.

iii. Provides a check facility for business relevance in that the user visually reconciles this table to the Test Rig, Transform, Match and Model Diagrams.

iv. Provides a communication facility because both business managers and analysts use the Business Rules Diagram as a common a tefact for discussion.

ii. Match Diagram COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:52 Wa l ington-Dummer Sydney No.1634 P. 23 matcn UtitledDiara4 YS. sQL Northwind products .avGQL Northwiri ~Prductfml categories u5up $erI D .Mat( 4 Matc Sal (AND)

C)

Co ii.

L Descrpion iuntPrice FPicture untsIntock UnitsOnaner PgReorderLavel f Dcontinued Collection of data tables networked tcgether to form a conglomerate data table. The user develops it by dragging tables from libraries or other diagrams.

Agile Data Mining® software executes this by translating it into a database query, carrying out the query, and returning the result as a conglomerate data tabli. The software performs the diagram-to-query translation internally, without user intervention. The software can execute match operations on tables from different databases or other data sources, ie. it can integrate data from disparate sources into a single table- Used for collecting and joining all the data that the user wants to use fcr the particular problem. In turn this can e used as an atomic Element in other diagrams.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:52 Wallington-Dummer Sydney No.1634 P. 26 24 0 iv. Provides an intuitive view of the data as Stables joined by relationships.

iii. Transform Diagram

VO

rt'-L Norl. lnda lrhoose 68ld,. geneate roaws, n i. A sequence of Elements representing a series of operations. The user develops it by dragging Elements and tables from libraries or other diagrams. At least one data table must be included to deliver a result, but an Activity Diagram without such a table can still be linked into other Test, Match, Model or Transform Diagrams.

ii. Agile Data Mininge software executes this series by translating it into a nested database query, carrying out the query, and returning the result as data tables or visual charts. The software performs the COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:52 Wal ington-Dummer Sydney No.1634 P. 27 25 o diagram-to-query translation internally, OC without user intervention.

Siii. Used to execute technical data mining l operations. In turn this can be used as an atomic Element in other diagrams.

Siv. Test Rig Diagram

VO

VO

i. Diagram to test either an appropriate single o Element or an appropriate single Technical Operation Diagram. It has four parts: a. Input data, which is a sample to be processed by the Element.

b. Element under test.

c. Expected output data, manually computed and entered by the user, or copied and pasted.

d. Actual output data.

ii. Agile Data Mining® software executes the test by running the Element with the given input data and comparing the actual output data produced against the expected output data.

Differences between actual and expected outputs are reported to the user.

iii. Test Diagrams are used only for checking that operations work as intended. (Checking COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:52 Wa l inston-Dummer Sydney No.1634 P. 28 26 o accuracy of data mining algorithms is an c analysis task, not a testing task, and is performed in technical operations diagrams as per other operations.) This ciagram provides assurance of work for both technical correctness and compliance to

O

business rules 0 v. Result Diagram i. A graphical display of data tables, predictive models, and visual charts that are computed by Transform or Match Diagrams.

It is generated on the same document that contains the corresponding Transform or Match diagram.

ii. Dependant on the output type, the user can interact with these outputs in a variety of context-sensitive ways. Eg- Highl.ghting table rows or zooming into regions on charts.

iii. Results are used for visual interpretation and analysis, as well as being an integral part of the reporting mechanism for the project.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:52 Wallingion-Dummer Sydney No.1634 P. 29 27 0 C vi. Model Building Diagram

C)

Co l i. A diagram to construct predictive nodels.

Data is input either by dropping a database Stable or by linking a Match or Transform

VO

\O Diagram. The user can select from several industry-standard data mining precictive o algorithms and set various parameters to ci control the model building process.

ii. Using the input data and the set parameters, this Diagram will generate a Predictive Model Element and statistics relating to the model's accuracy. Predictive Model Element generation is an industry-standard algorithmic learning process.

iii. The user drags the Predictive Model Element into a separate Transform Diagram to make predictions on other unseen data.

To execute a diagram, select its name from the Execution drop down list on the document construction toolbar and press the "execute" button (an arrow much like a media play button). This will then cause the software's engine to translate and run the diagram, linking in other COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 P- 2007 16:53 Wal ington-Dummer Sydney No,163C P. 28 0

C)

diagrams as required, and to generate any results on the document.

f. Templates contain skeletal outlines for common bu situations. The user drags a template onto a docume fills in the blanks with their own data. Once document the template can be altered to suit situations. Templates will be made available different industries and different business processe Libraries a. Libraries contain icons which represent Elemen used to draft each Agile Data Mining® document.

siness nt and on a custom for s.

i. The Elements within a Library are displayed within one or more groups. Each group is shown COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:53 Wallington-Dummer Sydney No.1634 P. 31 29 o using the title and frame as above. The group 0 can be expanded as shown or collapsed to display the title only. The groups are stacked Co In vertically within the Library.

ci ii. The following libraries are available Va i. Project Overview contains Elements for 0 odiagrams and documents already drafted for ci the current project.

ii. Standard Elements contains Element s for the standard building blocks that ship with the software- They represent data mining operations and business activities.

iii. Custom Made Elements contains Elements for building blocks custom made by the user that can be utilised in future work, iv. Data Sources contains Element for database tables, spreadsheets and files where source data resides.

v. Templates contains Elements for preassembled solutions for common business problems that the user can customise.

vi. Clipboard an empty space that can e used to temporarily store work.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:53 Wallington-Dummer Sydney No.1634 P. 32 30 o iii. To use an Element, drag it from the appropriate 0 Library to the Document being constructed. The software creates either a new copy on the Document or a link to the original instance as appropriate. The user can specify whether to ci link or copy some types of Element in certain Va situations.

iv. The user can save a Diagram as a reusable 0 oElement by dragging it from the Document to the Custom Made Elements Library. This creates a copy of the Element in that Library, and is available for use in the usual manner.

v. Libraries can be searched. A search box is contained at the top of the library, where the user can type text to match against. There are also several options to determine criteria: how to match the text and what to match it a ainst.

Erasing the text from the search box m kes it inactive. Typing text or clicking the options make it active; the search is recalculated after every keystroke or click.

lTfrCaira COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:54 Wallinston-Dummer Sydney No.1634 P. 33 31

C)

Co Process Five considerations may be taken into considerat solving a data mining subtask according to the invent: Business Rules: these are the criteria th subtask must meet and are formulations o characteristic of interest sought in the Data.

Test Rig: a testing arrangement to indicat the subtask meets the criteria set b Business Rules and is correctly implemente tests include data samples able to satis criteria and allow verification of c execution of the Technical Operations.

Activity Sequences: these are Tec Operations performed by the software in a of steps, for example as a seri calculations.

Relationships between tables: Tec Operations are arranged as a set of table relationship connectors between them; example field A of a table X is equal to I of a table Y.

Results: output from the Technical Operatd integrated into the Document to sho ion in .on; at the E some Source e that y the d. The Ey the orrect hnical series .s of .hnical s with for ield B ons is w the COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:54 Wallington-Dummer Sydney No.1634 P. 34 32 o eventual outcome. A visual inspection is ade to 0 ensure that the expected outcome of the Test Rig sample data has been correctly achieved Dy the o In3 subtask process.

0 Data Mnin Diagrams Each of the above considerations is met b five

VO

associated diagrams 44 as illustrated in Figure 4. These ci are a permanent part of the parent Document whilst this is oresident on the computer system, but may be export d for subsequent use, for example as embedded in repor s for business communication.

The Business Rules Diagram (Figure The Business Rules Diagram comprises a table 50 in which the user defines one rule per row. Columns provide details of the actual rule, the name of the rule, description and an example. A rule may be entered as text or as a Resource selected from one of the Libraries as described above.

The diagram provides a tool for planning and defining the scope of each Document. This gives fine-grain criteria related instructions, enabling accurate development of each Activity or Relationship diagram. It further provides a check facility of relevance to the set criteria of the Business Rules in that the user can visually reconcile the COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:54 Wallinston-Dummer Sydney No.1634 P. 33 o table to the Test Rig diagram and the Activity or C Relationship diagrams.

SThe Relationship Diagram (Fiqure 6) t This provides for a collection of tables 60 (database entries) networked together to form a conglomerate data table. It is constructed by the user by dragging tables

VO

VO from Libraries or from other diagrams linking them by relationship functions 62. The software of the system o translates the diagram into a query, executing it and ci returning the result in a conglomerate data table. The diagram to query translation is functionally performed by the software without user intervention.

The Relationship Diagram is used for collecting and joining all the databases that the -user wishes to interrogate for obtaining the solution to a particular data mining problem. It in turn can be used as a Resource in other diagrams. Furthermore it provides a visual intuitive view of the data tables and their connecting functional relationships.

The Activity Diagram (Figure 7) An Activity Diagram represents a series of operations and is developed by the user by dragging into the Document Resources and tables from Libraries and other diagrams. At least one data table must be included to return a result although an Activity Diagram without a COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:54 Wallington-Dummer Sydney No.1634 P. 36 34 o table may still be linked to Test Rig Diagrams, Relational CR or Activity Diagrams.

O The Activity Diagram is executed by the data mining

C)

VB system software after translation into a computation, returning the result as data tables, predictive models or visual charts. Again the software performs the diagram to IN computation translation internally, requiring no user intervention.

o The Activity Diagram executes the required data mining ci operations, with the derived output available for use as a Resource in other diagrams if desired.

The Relationship Diagrams and Activity Diagrams are characterised as the Technical Operation Diagrams if the data mining system.

The Test Rig Diagram This diagram is used to test either an appropriate single Resource or a single Technical Operation Diagram. It comprises four parts: input data; data structured by the user to reflect the type and characteristics of data to be retrieved by the data mining process, the Resource or Technical Operation under test, expected output data; this is manually derived by the user from the input data by applying the functions of the Business Rule as expressed in the Resource or Technical Operation.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:54 Wa I i ngton-Dummer Sydne No-1634 P. 37 35 o The system software executes the test by running the

C

Resource or Technical Operation Diagrams with the given Sinput data and comparing the result with the e pected Soutput data. Discrepancies between the actual output and expected output are reported to the user.

Note that the Test Rig Diagram is only used to assess

IO

IN the correct operation of the data mining system for z given problem. Actual checking of the accuracy of data mining o algorithms is an analysis task, not a testing task and is performed in Technical Operations Diagrams as per other operations. The Test Rig Diagram and its execution provide assurance of work for both technical correctne s and compliance with the criteria of the Business Rules.

Result Diagram A Result Diagram is a graphical display o: data tables, predictive models and visual charts that are computed by Technical Operations or Relationship Diagrams.

It is generated on the same Document which contains the corresponding Operations or Relationship Diagram.

Depending on the type of output, the user can interact with the Result Diagram in a variety of context felated ways. Results are used for visual interpretation and analysis, as well as providing an integral part of the reporting process for the data mining project.

Libraries COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:55 Wa I ington-Dummer Sydney No.1634 P. 38 36 o As noted above, Libraries contain a variety of 1 Resources represented by icons, which may be used to construct a data mining Document. With reference to Figure o 8, a Resource 83 from a Library 80 is accessed via icon 82.

The Resources within a Library are displayed within one or more groups. Each group is shown using the title 84 and

O

IO frame as shown in Figure 8. The group can be expanced (as in Figure 8) or collapsed to show the title only. Groups o may be stacked vertically within the Library.

ci COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:55 Wallington-Dummer Sydney No.1634 P. 39 37 o Available Libraries c1 Project Overview: contains Resources for diagrams and Documents already constructed for f the current project.

Standard Resources: contains Resources for the standard building blocks that ship with the data

O

NO mining software- They represent data mining operations and application activities.

o Custom Made Resources: contains Resources for building blocks custom made by the us r and saved for use in future data mining exerciles.

Data Sources: contains Resources for database tables, spreadsheets and files where Source Data resides.

Templates: contains Resources for pre-assembled solutions for common business and other application problems which can be customised by the user.

Clipboard: an empty data storage facility which can be used to temporarily store work.

To use a Resource, it is dragged from the appropriate Library to the Document under construction. The data mining software of the system creates either a new copy on the Document or a link to the original instance of the Resource as appropriate. The user is enabled to specify whether to link or copy some types of Resources in certain situations.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:55 Wallington-Dummer Sydney Nc-1634 P. 38 1^ 0 0 ci

C)

Co A user-constructed Diagram can be saved as a re- Resource by dragging it from the Document to the usable Custom Made Resources Library. This creates a copy of the Resource in that Library and is available subsequently in the normal manner.

In Use The data mining system of the present invention may be used in a variety of environments where data retained in various databases can provide bases for management decisions, if the various relationships and patterns inherent in the data could be extracted according to user defined criteria.

As an example, a sales and marketing department wishes to analyse its databases relating to its customers, to ascertain why the company is losing some customers while retaining others. The databases may include the customer list, sales databases and billing database, all maintained on the company's computer system server.

The object of the data mining exercise is to identify those customers the company is at risk of losing.

Typically, the user of the data mining software for a data mining exercise will be a data analyst who will work with sales and marketing staff and management to divide the objective into a number of loosely defined subtasks comprising smaller workable sections. These may comprise: Identify "all valuable customers" worth sa ing, COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:55 Wallington-Dummer Sydney No.1634 P. 41 39 o Identify those customers already lost :o the (C company, Create a profile for lost customers, V1 Match the lost customer profile against "all valuable customers".

Each of these subtasks may be addressed by the data

VO

DO mining software resident on the company's server. The first subtask, that of identifying all customers worth aving, o may be solved as follows.

ci The analyst creates a new Document by "clicking" a New Document icon on the toolbar of the user interface. The analyst, staff and management then determine the Business Rules to be applied to the subtask. These could be "find those customers who have made transactions greater than $1000 in the past year"; "find those customers who buy at least once a month"; find those customers who have made three or more transactions over the past 6 months" These Rules can be tabulated as shown in Figure 9.

The analyst now creates on the Document, one or more Test Rig Diagrams (Input Data and Expected Result Data) as shown in Figure 10, which contain sample data structured to satisfy the tabulated Rules. That is, if the Rules are correctly realized in the to-be-constructed Technical Operation Diagrams, the Input Data should yield the Expected Output Data shown in Figure 11- COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:55 Wallington-Dummer Sydney No.163 P. 42 40 o The formulation of the Business Rules and Test Rig

C

Diagrams may be an iterative process mediated between the

C)

analyst and management until both are satisfied that these n will capture the objectives of the data mining exercise.

The analyst now constructs the required Technical Operation Diagrams, (Activity and/or Relationship Diagrams)

OD

ID which implement the functionality set out in the Business Rules and Test Rig Diagrams. For this example, an Activity o Diagram addressing the first task "identify all valuable customers worth saving" would appear as shown in Figure 14.

The analyst now operates the data mining software to apply the Technical Operation Diagrams to the Test R g data to verify that the operations yield the correct expected outputs. If required, the Technical Operation Diagrams can be modified until the correct outputs are achieved.

Once satisfied that the Technical Operation Diagrams operate correctly on the Test Rig data, the data mining process on the organization's actual customer databases can be initiated with confidence that the output thus obtained conforms to the object of the exercise. The resultant output may take the form of tables, charts or combinations of these.

A new Document is created for each of the remaining subtasks identified with suitable Business Rules, Test Rig Diagrams and Technical Operations Diagrams as described for the first subtask. The final data mining solution is the COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:56 Wallington-Dummer Sydney No.1631 P. 43 41 o combination of all the subtask Documents into a single 0

C

overarching Document that executes each subtask in )sequence. For this example such an overarching project V Document would coordinate subtasks as shown in Figure 12, with a final output in this example, taking the form of the Stable shown in Figure 13.

VO

_O Although the above description is set in a business context it should be noted that the data mining system of 0 o the present invention can be applied to other than business problems. Thus for example in an engineering application the "rules" may comprise various engineering outcomes such as tolerances and surface finishes to be achieved by, various methods and available machinery, or stress and performance characteristics of various materials for example.

Software Background The above described preferred embodiments may be implemented by suitable programming of data processing equipment as follows: The application is a graphical Java application. The software is distributed on CD and installed on the user's computers. No internet-based service is provided as part of the core software.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 Se P 2 0 D7 16 5 Wallington-Dummer Sydney No-1634 P. 44 42 1^ 0 0 ci

C)

Co 0 The application communicates with external dat using standard software connectors.

The application communicates with databases ove network. The user does not interact directly wil database. Data analysis is executed by the applicat The business process is tightly integrated wit software's capabilities.

The application is installed on the desktop compu each user.

In alternative forms at least some of the sc components can be provided as embedded firmware on r built circuit boards.

The above describes only some embodiments

C

present invention and modifications, obvious to skilled in the art, can be made thereto without del from the scope and spirit of the present invention.

abases r the :h the ion.

h the ftware urpose if the those ,arting COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05

Claims (23)

1. A method for data mining of at' least one datal means of computer-implemented software; said including the steps of: a) creating at least one task defining Document f< of said at least one task, b) defining within said Document a Business Rules for said at least one task, c) defining within said Document at least one Te Operations diagram for implementation of B Rules of said Business Rules diagram, d) defining a Source Data icon indicating locat said at least one database or data file, e) executing said Technical Operations with said Data to generate at least one output diagram, f) verify that said at least one output complie said Business Rules; and wherein each said diagram is a graphical re constructed by means of said computer-implemented so each said diagram convertible to executable code for processing by a central processor of a computer s

2. The method of Claim 1 comprising the further defining within said Document data for a Te diagram to satisfy said Business Rules. ,ase by method )r each diagram -hnical isiness ion of Source s with adition ftware; adapted ystem. itep of st Rig COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:56 Wal ington-Dummer Sydney No.1634 P. 46 44 o

3. The method of Claim 1 or 2 comprising the further step 0 C( of verifying correct functionality by application of Ssaid at least one Technical Operations diagram to said C) iV data of said Test Rig diagram,

4. The method of any previous claim wherein said Document is composed by means of a user interface display VO ID generated on a display device linked to said computer and wherein descriptive and annotative text sections o may be defined with said document. ci

5. The method of any previous claim wherein said interface display comprises at least, a Document construction region, a Resource library region and a common productivity accessory region.

6. The method of any previous claim wherein said Eocument construction region is adapted to accept a combination of text and "drag and drop" Resources accessed from said Resource library area.

7. The method of any previous claim wherein one or more Resources are combined into a diagram in said Document construction region; each said diagram representing a subtask.

8. The method of any previous claim wherein at least one said diagram is a Business Rules defining diagram.

9. The method of any previous claim wherein at least one said diagram is a Technical Operations diagram.

COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:57 Wallinston-Dum er Sydney No.163 P. 47 45 The method of claim 9 wherein a said Technical r- Operations diagram may comprise an activity diagram, a 0) relationship diagram or a combination of activity and Co relationship diagrams.

11. The method of claim 10 wherein a technical operation C diagram may link in other technical operations diagrams \O \O which will embed and execute together when the former Sis run. o

12. The method of any one of claims 2 to 11 wherein said Test Rig diagram comprises a sample of input data and a sample of output data; said input data and said output data adapted to verification of one of said Business Rules and/or validation of one of said Technical Operations diagrams.

13. A computer-based data mining system wherein data mining is performed according to at least one user-defined rule for at least one associated data mining task; said system including a rule testing process wherein a sample of input data and a sample of expected output data are adapted to said at least one rule; aid at least one rule implemented through a Document based diagram structure wherein each of at least one diagram of said diagram structure is translated nto a computational process by said system. COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:57 Wal ogion-Dummer Sydney No.1634 P. 48 -4

14. The system of claim 13 wherein a said user-defined rule ciis a formulation of a characteristic of interest sought O in Source Data for a data mining operation.

The system of claim 13 or 14 wherein said system S includes construction of Technical Operations diagrams; said diagrams including relationship and activity Va diagrams.

16. The system of claim 15 wherein said relationship o diagrams represent a user-defined relationship between sets of Source Data.

17. The system of claim 15 or 16 wherein said activity .diagrams represent user-defined processes applic able to said sets of Source Data.

18. The system of any previous claim 13 to 17 wherein each of said diagrams is constructed by a user in a Document; said Document provided as a user interface on a computer display.

19. The system of claim 18 wherein said document is a readily interpreted corporate record of the busimass and technical steps involved that may be discussed, annotated, archived, reviewed, revised within the business operations. The system of any one of claims 13 to 19 where n each said diagram is translated by software of said data mining system into executable code for processing.

COMS ID No: ARCS-i 59837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:57 Wallington-Dummer Sydney No"1634 P. 49 47 Co 0\

21. The system of any one of claims 13 to 20 where user interface includes Libraries of Resources Resources including data mining operation application activities.

22. The system of any one of claims 13 to 21 where: user interface included productivity accessories accessories including calculator, a database dia tool and statistical functions.

23. A method for data mining of at least one datal .0 means of computer-implemented software; said including the steps of: creating at least one task defining Docum each of at least one task, defining within said Document a Business diagram for said at least one task, defining within said Document data for a T< diagram to satisfy said Business Rules, defining within said Document at leas Technical Operations diagram for implemental 0 Business Rules of said Business Rules diagra verifying correct functionality by applicat said at least one Technical Operations diac said data of said Test Rig diagram, defining a Source Data icon indicating loca said at least one database or data file, said said and .n said said gnostic ase by method nt Os t for Rules t Rig one on of on of am to t .i n, ic Ir tion of COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05 2007 16:57 Wallington-Dummer Sydney No.1634 P. 48 S executing said Technical Operations with said C-I Source Data to generate at least one output Sdiagram, Co ln verify that said at least one output complies with said Business Rules; Ci and wherein each said diagram is a graphical rendition VO \O constructed by means of said computer-implemented c software; each said diagram convertible to executable o code adapted for processing by a central processor of a computer system. COMS ID No: ARCS-159837 Received by IP Australia: Time 17:05 Date 2007-09-05