Association and Sequencing Products
DBMS, Data Mining Solutions Supplement
Decision trees are among the most common data mining techniques available today. Some products are suites consisting of several data mining tools; others are standalone products. All products (unless otherwise noted) include tree visualizers and generate predictive models which can be applied to a vast array of classification and regression problems.
* Angoss Software International Ltd: KnowledgeSeeker (Windows 3.x/95/NT, Unix) is a standalone decision tree product that features automatic as well as user-controlled tree growth and pruning. KnowledgeSeeker supports classification directly; regression requires binning of the continuous target variable. It offers choice of several split criterion and stopping rules.
For contact information see the product chart on page XX.
* Business Objects: BusinessMiner (Windows 95/NT) is an optional add-on to the Business Objects OLAP tool. It is descriptive model only. (It cannot be used as a predictor.)
* Cognos Inc.: Scenario (Windows 3.1/95/NT) is a CHAID-based decision tree implementation that builds models for descriptive purposes only (cannot be used for prediction). Intended for use in conjunction with Cognosıs OLAP tools.
* HyperParallel Inc.: //Discovery (Unix on Sun, Pyramid and NCRıs Worldmark) includes a CART-based tree engine, //Induction. Supports both classification and regression, detecting the type of model to produce based on the datatype of the target variable. Limited built-in presentation capabilities with tree output available only in text format. Algorithm is scalable on parallel processing architectures.
* Integral Solutions Ltd.: The Clementine toolset (Unix, VMS, Windows NT) includes two decision tree algorithms, C5.0 for classification and an ID3 based tool for regression. Output is limited to rulesets only (no tree visualizer). Simple and expert modeling levels are supported with access to numerous modeling parameters.
* IBM: The Intelligent Miner suite (AIX, OS400, MVS) includes a decision tree for classification. Control over modeling is limited to minimal pruning parameters. The algorithm is scalable on parallel processing architectures.
* NCR: Knowledge Discovery Workbench is an implementation of Integral Solutions Ltd.ıs Clementine. See details above.
* NeoVista Software, Inc: Decision Series (Unix on DEC, HP and SUN) includes a C4.5-based implementation as its decision tree component, DecisionTree. Supports classification only. Trees can be produced in batch or iterative mode, the latter allowing for automatic production of multiple trees and, optionally, rulesets. Tree and rule presentation is text only. Its algorithm is scalable on parallel processing architectures.
* SAS Institute Inc.: SAS Enterprise Miner (Windows 3.1/95/NT, OS2, Macintosh, any Unix, IBM MVS/VM/VSE, DEC VMS). SAS Data Mining Solution, an optional add-on to the SAS system, includes a CHAID-based decision tree component for classification.
* SPSS Inc.: SPSS-CHAID (Windows 3.1/95/NT), an optional SPSS tool, allows users to generate models useful only as descriptive models.
* Silicon Graphics Inc.: MineSet (Silicon Graphics Indy/OS2/Indigo workstations, Challenge and Origin servers). MineSet supports classification through its Decision Tree Inducer. MineSetıs decision tree implementation most distinguishing characteristic is its 3D color tree visualizer which features a ıfly-overı capability that allows the user to interactively explore the tree.
* Thinking Machines Corp: Darwin (Unix on IBM RS6000, Sun). StarTree, Darwinıs decision tree component, uses a CART-based algorithm to produce predictive models for classification and regression. Many tree growth and pruning parameters. Pruning produces multiple subtrees from which the user can choose the best. Display is text only. Algorithm is scalable on parallel processing architectures.