XpertRule Miner

Category Intelligent Software>Data Mining Systems/Tools

Abstract XpertRule Miner is a graphical data mining system that can be used to discover tree patterns from historic business data. Using ActiveX technology, the Miner client can be deployed in a variety of ways. Solutions can now be built as stand-alone mining systems or embedded in other vertical applications under MS-Windows. Deployment can also be over Intranets or the Internet. The ActiveX Miner client works with XpertRule's high performance data mining servers to provide multi-tier client-server data mining against very large data bases. Mining can be performed either directly against the data in situ, or by high performance mining against tokenized cache data tables.

Miner includes extensive data transformation, visualization and reporting features. Data can be manipulated using a drag and drop interface. Users can graphically design their customized data manipulation, mining and reporting processes. Software developers can also directly control the application using the exposed methods and properties of the Miner's objects. This enables Miner to be seamlessly integrated as part of vertical applications - such as for a Customer Relationship Management system, which could have been built in any environment [Visual Basic (VB), Delphi, etc.]. All this can be achieved without compromising scalability or performance.

Data can be filtered through tree profiles and reported on, or extracted to new data sources. Graphs, charts and tables can be dynamically linked to the graphical tree views of your data. As the user navigates through different levels and applies different scenarios they can be kept constantly updated.

Data Transformation - Using a drag and drop interface, data transformations can be designed and executed. Interim reports can be included to monitor progress, using expand/collapse data structure tables and 3D data exploration graphs.

Table Manipulations - 1) Table joins - The data fields from different tables can be joined together according to a common index field; 2) Table merge - The rows from different tables with the same field definitions can be merged; 3) Table filtering - Rows can be excluded from tables according to a given criteria. They can also be filtered through trees to find out which profile they belong to;4) Table sorting - Tables can be sorted into ascending or descending order; 5) Table sampling - Random samples can be extracted from larger tables.

Field manipulations - 1) Derived fields - New fields can be generated from calculations or string manipulations on existing fields. This can be used to clean data fields or to generate new fields for data mining. Features drag and drop of field variables and Visual Basic like syntax' 2) Aggregated fields - New fields can be generated by aggregating the data in other data fields to generate summaries such as minimum, maximum, average, sum and count. Time series aggregation can also be performed.

XpertRule Miner SDK - The Miner's System Developer Kit (SDK) can be used for deploying embedded Miner OCX (OLE Control EXtension) tree components and data transformation scripts. The kit includes templates and examples for Visual Basic and Delphi, and generation of default HTML (Hypertext Markup Language)/script pages for Web deployment.

ProfilerX - ProfilerX is the ActiveX decision tree mining component included with XpertRule Miner. It can be used as an object within the Miner's graphical mining process. The exposed properties and methods of ProfilerX also enable it to be seamlessly embedded as a data mining component in any application by using the Miner's SDK.

ProfilerX does Not use memory loading [i.e. Random Access Memory (RAM)] to mine data. All mining is performed using client-server either to remote data [i.e. using Structured Query Language (SQL) to mine the data on the server] or to the data on disk using high performance compressed table mining (called Data Mining Tables or DMT). ProfilerX many features include -

• 1) Can mine very large numbers of table rows (millions) without sampling;
• 2) Horizontal or vertical tree orientation with tree magnification zoom, secondary tree overview and navigator window;
• 3) Synchronized tree path display (expressed as English text) with path highlighting;
• 4) Tree induction on discrete or numeric outcomes;
• 5) Automatic tree induction, interactive tree induction (user override), or manual tree branch editing using drag and drop;
• 6) Ability to create a decision from one outcome, then change the outcome used on the tree and see how that outcome's frequencies fall down each path;
• 7) Automatic and manual field grouping. Numeric attributes can have automatic thresholds derived or used "as is";
• 8) Field ranking on up to 1000 fields per pass;
• 9) Multiple secondary report windows with graphs and tables for field frequencies, probabilities, aggregated summaries, cross leaf and lift/gain charts;
• 10) Secondary report windows can be updated dynamically to redraw graphs and tables as the user navigates around the decision tree;
• 11) User controlled color/intensity settings for frequencies and probabilities;
• 12) Normalization of frequencies;
• 13) Copy and paste of decision trees as WMF (Windows Meta Files). When pasted into MS-Word or MS-Excel these can be edited and annotated;
• 14) Monitor multiple data sets by adding them to existing trees. Visualize the differences between monitored data sets using graphs;
• 15) Decision trees can be used to filter records (table rows) as part of XpertRule Miner operations.
• 16) Multiple tree objects and tree filters as part of XpertRule Miner operations;
• 17) Requires the XpertRule Miner SDK for deployment.

Associations Discovery - XpertRule Miner supports the discovery of associations in both "case" and "transaction" data.

Case based data is the type typically used for tree induction in which each record (or table row) has data fields describing that particular case - for example, a customer record. In contrast to tree induction which needs an outcome field to be defined, associations discovery can discover inherent clusters/segmentation in case data. This can be used to provide better understanding in case data or to define new data fields (i.e. table columns) prior to tree induction mining.

Transaction data consists of records each representing a transaction identifier and an associated item or event. Each transaction will appear in as many records as there are items associated with it. The discovery of associations in transaction data is typically used in "basket analysis - to discover affinity between purchased items.

Note: Mined trees can also be exported to XpertRule Knowledge Builder (see G6G Abstract Number 20031U9).

System Requirements

• Microsoft Windows Operating Systems
• At least the minimum memory recommended by the operating system.
• ODBC - Open Database Connectivity.

Manufacturer

• XpertRule Software
• Innovation Forum
• Innovation Park
• 51 Frederick Road
• Salford
• M6 6FP
• UK
• Email: info@xpertrule.com
• Tel: +44 (0)870 60 60 870
• Fax: +44 (0)870 60 40 156

Manufacturer Web Site XpertRule Miner

Price Contact manufacturer.

G6G Abstract Number 20032

G6G Manufacturer Number 103005