MineSet implements a client-server architecture. The client and the server need separate licenses. Typically, a client is used by a single user on a desktop system, while the server runs on a larger system that may be shared by multiple clients simultaneously. The client side runs the Tool Manager and the visualization tools, and the server side runs the DataMover and analytics. You may run the client and the server on the same system, but you need both a client license and a server license.

Client licenses are simple. There is only one type of client license, and it is tied to a system ID. Once you have a MineSet client license on your system, you may run an unlimited number of visualizers and tool-manager processes on that system as long as they read local files and do not connect to a server.

Server licenses are more complex. In essence, each server license allows one simultaneous connection to the MineSet server. For example, if you have five users working simultaneously, each using one client, you need five server licenses on the server system. Unlike a client license which is unlimited on one system, a server license means support for only one active session.

There are two types of MineSet server licenses:

Normal server licenses are priced in such a way that they become cheaper the more licenses you buy. As of MineSet 2.6, there are three tiers of Normal server licenses:

Lastly, there is a special kind of inexpensive server evaluation license called “MineSet Light.” If your dataset has fewer than 5,000 records, MineSet uses all of them. If your dataset has more than 5,000 records, MineSet auto-samples to 5,000 records and uses only those.

Mixing Varsity with any other license type is not allowed. Mixing Normal with Light licenses is not recommended. In MineSet 2.6, exceeding the number of licenses on the server generates a warning. The warning alerts you to the fact that you are exceeding your server licensing terms. Stricter control of license usage may be introduced in the future.

Beginning with version 2.6, MineSet supports international datasets. Text labels in the graphical interface still appear in English, but you can now view multibyte column names and data values in the language corresponding to the data encoding. MineSet automatically supports EUC encoding for Japanese, Chinese, and Korean, provided the corresponding WorldView product is installed. For other languages and encodings see “Extending to Other Languages and Encodings”.

locale -a 

setenv LANG ja_JP.EUC 

Large memory (64-bit) is supported on IRIX 6.4 and later releases. If you have IRIX 6.2, you can still use the 32-bit data mining utility, but you must upgrade to IRIX 6.5 in order to obtain 64-bit support and pthreads. To get the full advantage of 64-bit addressing you may also need to change the systune resource parameters, depending on your system configuration.

The systune parameters determine the default limits on the available system resources. Table 1-3 lists the systune parameter values that Silicon Graphics recommends (for more details see the systune(1M) man page):

---

Note: You must reboot your machine after installing the new parameters.

---

MineSet now supports Y2K-compliant dates. In the U.S. locale, dates may be entered in the form MM/DD/YY or MM/DD/YYYY. MineSet follows the X/Open standard for two-digit years: numbers greater than 68 are assumed to be the years 1969 to 1999, and numbers less than or equal to 68 are assumed to be the years 2000 to 2068.

In European locales, dates may be entered in the form DD/MM/YY or DD/MM/YYYY, with the same handling of two-digit years as above.

In either locale, if you enter a two-digit year, it is automatically expanded to a four-digit year in the display.

Now, with the use of an API, third-party vendors can extend the functionality of MineSet 2.6.

The MineSet Mining Tool Plugin API provides a means for third parties to plug in a GUI to the Mining Panel in the Tool Manager, save options to a .mineset file, and send these options to the DataMover. The DataMover then runs the third-party mining tool program and manages its output of models and model visualization files. The model visualization files are then sent back to the Tool Manager, which runs the requested visualization. Models created by third-party plugin mining tools can later be applied to a dataset using the Apply Model transformation in the Tool Manager.

At startup, MineSet looks for third-party dynamic shared object (DSO) libraries in /usr/lib/mineset/plugins.

In MineSet 2.6, a clustering algorithm named AutoClassPro (ACPro) from Ultimode Systems is available as an add-on. If ACPro is installed, documentation for it can be found in /usr/acpro/doc.

dataschema is a MineSet data-importing utility that automatically creates MineSet data and schema files from flat file formats.

dataschema /tmp/mydata.csv 

In MineSet 2.6, bin names have a new format. MineSet 2.5 used the convention - 10, 10-20, 20-30, 30+ for bin names, which led to some confusion. The - 10 was often thought to be negative ten, and it was not clear which bin contained the boundary points. MineSet 2.6 uses a modified interval notation for bin names:

(lower-bound ... upper-bound]

The “(” indicates that the lower bound is not included in the range. The “]” indicates that the upper bound is included in the range. For example, (10.5 ... 12.6] indicates the range of values over 10.5 up to and including 12.6, more formally, { X : 10.5 < X <= 12.6 }. If the lower bound is omitted, the range includes all values less than and including the upper bound. For example, (... 10.5] indicates the range of values less than or equal to 10.5, or more formally, { X : X <= 10.5 }.

If the upper bound is omitted, the range includes all values greater than the lower bound. For example, (12.6 ...] indicates the range of values greater than 12.6, or more formally, { X : X > 12.6 }.

The example of the MineSet 2.5 bin names - 10, 10-20, 20-30, 30+ can be expressed in MineSet 2.6 as (... 10], (10 ... 20], (20 ... 30], (30 ...]. Other examples make the naming scheme clear:

The Histogram Visualizer automatically bins all of the continuous-type columns of in the data and sends the result to the Statistics Visualizer. Figure 1-1 shows the following Histogram Visualizer options:

The Scatter Visualizer now allows you to show a trail of motion to demonstrate the changing animation path of an entity. When you create an animation, the trail shows behind each selected entity, in the form you have selected. The motion option menu, located at the bottom right of the ScatterViz control panel, allows you to select from:

All trails are color-coded according to the originating entity. If an entity changes from red to blue as the summary slider changes from one position to another, the corresponding trail will also be shown changing color gradually between the two positions. Trails are made between points whose unmapped attributes stay the same over the course of the path.

Aggregated data grouped by a small number of columns tends to be an excellent candidate for the display of motion trails. Initially, motion trails are displayed for all points in the scatterplot affected by the path. Selecting any entity by clicking on it with the mouse causes only the selected point to display a trail. This can be used to reduce visual clutter. Entities with null positions appear as breaks in the trails.

Figure 1-2 shows an example of the Scatter Visualizer with tube motion trails.

Figure 1-2. Example of ScatterViz Tube Motion Trails

To support the new association rules visualization capability of the Scatter Visualizer, the following statements have been added to the Scatter Visualizer configuration file. Refer first to Appendix D, “Creating Data and Configuration Files for the Scatter Visualizer,” in the MineSet User's Guide to understand the basic format of the Scatter Visualizer's configuration file.

disk height clause1, clause2,... 

disk color clause1, clause2,... 

drillthrough var 

axis LHS, orderby alpha; 

The Decision Table Visualizer has two new features: filtering and Evidence mode.

Evidence Mode in the Decision Table Visualizer

The Decision Table Visualizer now has an Evidence mode that is the same as the Evidence mode of the Evidence Visualizer. Go to the View menu of the Decision Table, and choose Evidence Mode to activate it.

The distribution in each cake now shows conditional probabilities, rather than distributions based on the record weights, which are shown initially. This is useful if one of the classes is small.

Figure 1-3. Decision Table Breakdown Showing Distribution Based on Race

In Figure 1-3, the label is “race” and the distribution of the data is based on weight. Because most records in this dataset are labeled race=white, it is difficult to discern what values give evidence for other races. Switching to Evidence mode (Figure 1-4) makes it clear which regions give evidence for races less prevalent in the data. See “Selecting Items in the Main Window,” in the MineSet User's Guide for the technical description of how evidence is computed.

Figure 1-4. Decision Table Showing Conditional Probabilities

Normalized conditional probabilities (evidence) are shown at each cake when using the Evidence mode. From the visualizer window pull down the View menu and choose Show as Evidence.

In Figure 1-4, if the racial breakdown at a cake matches the prior probability (shown in the Label Probability window at the right of the view area), then the slices are of equal size. Bigger or smaller slices indicate correspondingly more or less evidence for a given race. If the slices for a cake are all of equal size, then the racial breakdown for that combination of values is the same as the prior distribution (the distributions shown originally in the pie in the right window.)

The Decision Tree Inducer now has an extended set of splitting criteria and pruning methods.

All visualization tools now provide the option of creating a file based on a visualization that may be published on the Web. From the File menu, select the “Publish on the Web” option. This runs a script that produces a .mtr file which is placed in a selectable directory. The default directory to appear in the file selection dialog box is one of the following:

There have been a few changes to the file exchange procedures between MineSet and SAS. The following sections describe the changes or replace equivalent sections in Chapter 9, “File Exchange Between MineSet and SAS,” in the MineSet User's Guide:

setenv SAS_CMD /usr/people/joe/sas/bin/sas  (IRIX) 
set “SAS_CMD=c:\Program Files\sas\sas.exe”  (Windows NT)

mineset2sas <MineSet file> <SAS datafile> [options]

mineset2sas cars cars.ssd01 -svsc -names cars.names

sas2mineset <SAS datafile> <MineSet file> [options]

sas2mineset houses.ssd01 -svsc -names houses.names