There are five new nodes in WITNESS Miner v2 to support different stages of the KDD roadmap. Click on each of the links below to find out more about these new features.
The file node from version 1 has been renamed the data
source node. The node now supports loading databases via OLE DB
data sources. Drivers compatible with OLE DB include Microsoft Jet,
ODBC, SQL Server, and any other data source that can represent data
in a tabular format and has a suitable provider.
The data source node now allows categorical fields
to be identifed from within the program rather than using the
asterisk notation in the datafile (see data
source node).
The speed of loading databases has been improved in
the data source node and the ability to abort a load operation is now available.
The distribution node
has been updated to that the distribution of the target classes can
be identified within each bar on the graph.
The rule node can now
display results to the screen in standard format or using SQL query
format (see Rule Format below).
Significant improvements have been made to the discretisation node. The boundaries from the discretisation algorithms are now displayed in the log window. This information can be used with the new manual discretisation option to reproduce a particular partitioning across fields in other database, for example, to apply the same discretisation to a training and testing dataset. Setting options for discretisation is now much simpler using the discretisation wizard.
Reports and results dialog boxes have been improved to use the new HTML report generator. As a result, reports have a common look and feel to them as defined by the standard WITNESS Miner style sheet. Reports can be readily exported in HTML format with style tags that can be easily to personalise the report style for your intranet.
Improved HTML reports can be found in the following nodes:
The rule node is now capable of exporting rules in XML format in addition to the standard WITNESS Miner format. The XML rules can be used in web browsers to generate reports or just to view the raw data. Alternatively, the rules can be loaded into most programming and scripting language and processed as a standard XML document structure.
Two XML style files are provided for formatting exported rules when using a web browser: dlwrules.xsl and sqlrules.xsl.
The exported rules also conform to the RuleSchema.xml schema file to ensure the data is well-formed.
The options dialog box has a fourth tab: Reports.
These options can be used to select which style sheet will be used
to build the report.
Double-clicking a node on the toolbar will
automatically drop a new node of that type onto the centre of the
workspace. See quick create
for more details.
Unlink Mode. Links in a stream can not be unlinked
(and relinked) whilst retaining their settings. Full details on the toolbars page.
Node Cloning. Nodes can easily be duplicated
(including their data and options) by right clicking on a node and
selecting 'Clone' from the context menu. For further information see cloning
nodes
With the increase number of nodes in version 2, the
nodes toolbar has been divided up into six smaller bars relating to
the six supported stages of the KDD roadmap.
The rule scratch pad can now be viewed from the Edit menu.