A b o u t    V i s i R e x
Table Of Contents

About VisiRex

License Agreement

Specifications

How to Purchase

Version History

Future Plans



VisiRex Specifications

  • VisiRex is based on a highly optimized version of the C4.5 algorithm published by J. Ross Quinlan.
  • VisiRex is a 32 bit program which requires Windows 95, 98, ME, 2000 or Windows NT4(sp3).

Data Requirments

  • Data must be in a table contained within an Access MDB file, up to one gigabyte in size.
  • VisiRex Import function converts any comma separated (CSV) text file to MDB format.
  • VisiRex can link to most common databases if you have their ODBC drivers installed.
  • VisiRex only reads your data file. Your data is never modified in any way.
  • Data may contain 2 to 255 fields and any number of records.
  • One data field must be set as primary key index.
  • Fields may be bit, byte, integer, floating point, date or text type.
  • Text fields selected for use by VisiRex may have up to 256 unique values.

Feature List

  • Ease of use has been the ultimate goal in designing VisiRex.
  • Data may be imported from ASCII text files.
  • You may select which fields are to be used by VisiRex rule extraction.
  • You may select whether numeric fields are used as continuous values or discreet classes.
  • You may select which rows are used for training the rule extraction, and which rows are to used for later testing.
  • Several VisiRex projects may share the same source database.
  • Rows containing missing values will automatically be detected and handled.
  • Extracted rules may be viewed in either flow chart format or tree format.
  • Extracted rules may be interactively pruned using a combination of two pruning algorithms.
  • Flow chart nodes may be color coded to indicate data volume, prediction confidence, or prediction trend.
  • Every point on the flow chart reports data volume and prediction confidence.
  • Flow charts may be tiled to your printer, or exported to Visio ®.
  • Comprehensive statisical reports are automatically compiled for your database.
  • Confusion matrix shows actual versus predicted for all classes.
  • Data may be subdivided, then browsed and exported according to position on the flow chart or confusion matrix.
  • Data browsing allows multiple sorting, text finding, and checkmark setting.
  • Checkmarked rows may be exported as a new MDB database or as ASCII CSV files.
  • Live predictions can be made, enabling you to interactively experiment with field values.
  • Context sensitive help file responds to your F1 key.

Speed Tests

  • These amazing speeds were achieved on a stand-alone 400 Mhz Pentium II computer using 128 Mb Ram and IDE hard drive.
  • The Cancer database contains 458 rows of data. Using 9 numeric input fields a rule tree consisting of 45 nodes was extracted in 1 second.
  • The Mushroom database contains 8,124 rows of data. The data took 17 seconds to initially load into VisiRex. Using 22 text input fields a rule tree consisting of 29 nodes was extracted in 1 second. Testing this tree for all rows required 15 seconds.
  • The States database contains 28,052 rows of data. The data took 5 seconds to initially load into VisiRex. Using 2 numeric input fields, a rule tree consisting of 225 nodes was extracted in 4 seconds. A unpruned tree consisting of 1,429 nodes was extracted in 6 seconds. Testing this 1,429 node tree tree for all rows required 11 seconds.
  • A manufactured database contains 1 million rows of data. The collection of data statistics at project creation took 10.5 minutes. The data took 6 minutes to initially load into VisiRex. Using 4 text fields plus 2 numeric fields, a rule tree consisting of 17 nodes was extracted in 18 seconds. Testing this tree for all million rows required 11 minutes.

A Complete System for Inductive Rule Extraction

CorMac Technologies Inc.
34 North Cumberland Street ~ Thunder Bay ON P7A 4L3 ~ Canada
E m a i l