Jump to: navigation, search

Typical Response Times

This topic describes part of the functionality of Genesys Content Analyzer.

It includes some typical response times for Genesys Content Analyzer. For other functions of Knowledge Manager see the discussion in Typical Response Times.

Unless otherwise stated, these figures are for a machine running Windows 2000 with two Pentium 4 processors and 1 GB of RAM.

  • Deleting a training object takes approximately 4 seconds per 1,000 e-mails.
  • Copying e-mails from one training object to another takes approximately 8 seconds per 1,000 e-mails.
  • Creating a model (training time) naturally varies with the number of categories, number of e-mails, selected training quality, and selected cross-validation. As one example, for a training object containing 76 categories and 73,000 mails, with training quality set to level 1 and no cross-validation, training time is approximately 29 minutes. This is on a host running Windows 2000 with one 600 MHz processor and 1 GB of RAM.
  • Cross-validation may increase training time significantly. The table "Increase of Training Time with Cross-Validation" shows, for selected cross-validation levels, the factors of increase of cross-validation over no cross-validation.

Increase of Training Time with Cross-Validation

Cross-Validation Level

Factor

3

1.9–2.9

5

4.0–4.7

10

8.0–9.0

    For example, training a model at cross-validation level 3 takes between 1.9 and 2.9 times as long as the same model with no cross-validation.
    • Classification performance depends on the size and nature (level of training quality) of the model. The table "Classification Performance" shows some examples, all of which use a test object that contains 3,726 text objects.

    Classification Performance

    Model

    Host Machine

    Classification Rate

    72,734 text objects
    Size = 285 KB
    Quality = 1
    Cross validation with split to three sets

    Two Pentium 3 processors
    512 MB RAM
    Windows operating system

    31 objects classified per second

    72,734 text objects
    Size = 309 KB
    Quality = 3
    Cross validation with split to 10 sets

    Two Pentium 3 processors
    1 GB RAM
    Windows operating system

    28 objects classified per second

    72,734 text objects
    Size = 309 KB
    Quality = 3
    Cross-validation with split to 10 sets

    Four 350 MHz processors
    4 GB RAM
    Solaris operating system

    15 objects classified per second

    • An FAQ object can process 30–50 classification requests per second on a model that contains 500–1,000 categories.
This page was last edited on April 28, 2014, at 18:32.
Comments or questions about this documentation? Contact us for support!