You are here: Foswiki>Software Web>MineSet (27 Jul 2009, ProLan)Edit Attach
MineSet - http://mineset.sgi.com

Mine Set 2.5 released in May 1998.
It is a fully integrated, comprehensive suite of easy-to-use
analytical and visual data mining tools. Mine Set 2.5 software tools
revolutionize customers' decision support process by offering
parallelized data mining algorithms for faster performance.


Visual data mining:
Mine Set enables interactive exploration of data
through an advanced suite of visual tools for
faster discovery of meaningful new trends and
relationships. The Splat Visualizer and the
Scatter Visualizer represent complex data in
up to eight dimensions. The Map Visualizer
displays data with strong geographical relationships
by using a map metaphor. Animation and view
synchronization techniques are used to reveal
patterns over critical dimensions such as time.
The Tree Visualizer depicts data with hierarchical
relationships utilizing a fly-through technique
set in a 3D landscape. The Statistics Visualizer
presents a visual summary of basic statistical
information. Advanced drill-through techniques
give you fast access to the original records that
created entities within your visualization for
additional exploration and analysis.


Analytical Data mining:


Below the elegant interfaces of the visual tools,
state-of-the-art analytic data mining algorithms
build comprehensive models. Mine Set includes
multiple classifier inducers including a Decision
Tree, Option Tree, Evidence, and Decision Tables.
Boosting allows Mine Set to further improve the
accuracy of these classifiers. Decision Tree and
Option Tree classification models are viewed
through the Tree Visualizer. An associated tool, the
Evidence Visualizer, displays the structure and
properties of the Evidence Classifier and supports
instantaneous what-if analysis. The Decision Table
Visualizer displays multidimensional data on
important columns automatically chosen by the
matching inducer. The spreadsheet-like paradigm
supports multilevel drill-through capability. The
Regression Tree inducer supports regression and
the resulting model is visualized using the Tree
Visualizer. The Association Rules Generator
analyzes data to discover product affinities and
relationships between data entities. The resulting
rules are depicted by the Rule Visualizer in an easy-to-
understand graphical format. The Column
Importance feature enables automatic or user-guided
selection of important columns for use with
the Mine Set visualizers. The Clustering algorithm
segments your data into similar clusters and is
visualized using the Cluster Visualizer.

Features:


o Data Access and Transformation Features
- Support for access to Oracle, Informix, and Sybase
- Support for a three-tier architecture to accommodate
heterogeneous computing environments
- Support for access to flat data files (ASCII or Binary) on
client or server GUI-based and SQL querying of RDBMS


o Data transformation history and graphical editing facility
o Data transformation support for:
- Automatic or user-specified binning of data
- Data aggregations with indexed arrays using average,
minimum, maximum, sum, and count
- Data distribution (transpose)
- New column creation using expressions
- Column removal
- Data type conversion
- Sampling
- Application of classifiers to existing data sets
Save and restore session management

o Client/server architecture
o Statistical tool for finding minimum, maximum, means,
median, standard deviation, histograms, and quartiles
o SAS file import/export utility
o Analytic Data Mining Features
- Classifier induction of decision trees, evidence
(Simple Bayes), decision tables.
- Boosting
- Clustering
- Regression
- Automatic accuracy estimation (holdout and
cross-validation)
- Automatic attribute discretization (binning)
- Integration with Mine Set visualization tools
- Laplace correction for evidence induction
- Automatic feature selection for evidence induction
- Data holdout percentage for accuracy estimation
- Scoring
- Record weights
- Loss Matrices
- Learning curves
- Lift curves
- Backfitting
- Association rule generation
- Automatic column importance selection


o Visual Data Mining Features
- User-defined mapping of data to visualization components
- Visual and numeric normalization of data across visual
components such as height, color, and size
- Discrete or continuous mapping of data to colors
- Data aggregation summary windows
- Web launching support
- Visual drill-up and drill-down of data
- Visual drill-through to source data
- Animation of dependent data by up to two user-defined
independent variables for trend analysis
- VCR-style animation playback
- Visual filtering and querying of data
- Automatic visual scaling adjustment
- Synchronous animation of multiple visualizations
- Visual level-of-detail controls
- 3D fly-through
- Pan, rotate, zoom, and dolly control for point of view
- Integrated execution of UNIX shell commands
- All visualizations are based on Open GL and can be viewed
and controlled from platforms running an Open GL enabled
X server. Mine Set includes a trial version of Hummingbird Communications
Exceed and Exceed 3D software, which run on PCs with
Windows(R) and Windows NT (R)


* Platform(s): Any Silicon Graphics platform running Irix 6.2 or higher.
Any PC running Hummingbird's Exceed 3D X-server can be
used to hook in as Mine Set client.


*Contact: mineset@sgi.com (preferred) or
Aydin Senkut, phone 650-933-6154

Contributed by: Ron Kohavi, mineset@sgi.com

Note: This info converted from the original "The Data Mine" pages and pre-dates June 2001. Please remove this note if you update or check the info
Topic revision: r5 - 27 Jul 2009, ProLan
 

This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Foswiki? Send feedback