Databases Reference
In-Depth Information
data scientists work with familiar tools and languages. That's a subtle yet powerful ca‐
pability of Lingual.
Books Related to Lingual
For more information about ANSI SQL issues related to Lingual, check out these topics:
Goodman, and Julian Hyde (Manning)
Pattern, a DSL for Predictive Model Markup Language
Pattern is an extension to Cascading that translates
Predictive Model Markup Language
(PMML)
into Cascading apps. This open source project is a collaboration between de‐
velopers at Cascading and other firms, to get coverage for several popular machine
learning algorithms.
PMML is an established XML standard, since 1997, developed by a consortium called
Data Modeling Group
.
Many vendors for analytics frameworks support exporting
models as PMML: SAS, IBM SPSS, Microstrategy, Oracle, etc. Also, many popular open
source tools support PMML export: R, Weka, KNIME, RapidMiner, etc.
PMML is the leading standard for statistical and data mining models and supported by
over 20 vendors and organizations. With PMML, it is easy to develop a model on one
system using one application and deploy the model on another system using another
application.
— Data Modeling Group
The XML captures the parameters of a model, plus metadata for defining it as a work‐
flow. That's the point of Pattern: develop models on popular analytics frameworks, then
deploy them within Cascading workflows. Benefits include greatly reduced develop‐
ment costs and fewer licensing issues at scale; leveraging the economics of Apache Ha‐
doop clusters, plus the core competencies of analytics staff, plus existing IP in predictive
models.
Organizations also like to use PMML for this work because several different models can
be trained, and then the resulting PMML gets tagged and archived in version control.
Experiments can be evaluated with
A/B testing
,
multi-armed bandit
,
etc.; however, the
source code does not have to change as the models evolve.
