Online predictive analytics

JPMML prediction processor evaluates event feature vectors in near-real-time

Objective

Provide streaming predictions and scoring using a standard PMML online predictor implementation.

The implementation leverages the JPMML open source library developed by Villu Ruusmann.

Example

The example below demonstrates the ease at which a machine learning model can be embedded within a stream processing pipeline and have events scored.

Model preparation

On Joule startup the following is performed before events are processed:

Model is loaded in to memory from disk.
The required feature vector attributes are derived from the model ready to be retrieved from the presented event.
A response field is configured.

pmml predictor:
  name: irisScorer
  model: ./models/iris_rf.pmml
  response field: flowerPrediction

Model scoring

Events are scored in near real-time using this process:

Extract the required feature vector attributes from the event.
Execute the model inferencing function using the prepared feature vector.
Place the resulting score in to the event assigned to the response field.

Attributes schema

Attribute Description Data Type Required

Attribute	Description	Data Type
name	name of the counter.	String
model	model filename that specifies path and model filename.	String
model store	Reference data logical store name where the model resides. See Contextual data documentation for further information on configuration	String
features field	Field where all the required features exist for the model prediction function. This is an optional field where engineered map of features are placed and applied to the model evaluation. See the Feature Engineering documentation for further details	String
response field	Name of field where the result of the prediction function will be assigned too	String
unpack results	Unpack the algorithm response variables directly into the `StreamEvent` object. Otherwise the process will add the complete response object in to the response field	Boolean Default: false
audit configuration	Every prediction can be audited along with the features used.	See model audit page

name

name of the counter.

String

model

model filename that specifies path and model filename.

String

model store

Reference data logical store name where the model resides. See Contextual data documentation for further information on configuration

String

features field

Field where all the required features exist for the model prediction function. This is an optional field where engineered map of features are placed and applied to the model evaluation. See the Feature Engineering documentation for further details

String

response field

Name of field where the result of the prediction function will be assigned too

String

unpack results

Unpack the algorithm response variables directly into the StreamEvent object. Otherwise the process will add the complete response object in to the response field

Boolean

Default: false

audit configuration

Every prediction can be audited along with the features used.

See model audit page

Support

For more details or assistance, Fractalworks is ready to support in your enquiries.

PreviousTransform NextModel audit

Last updated 5 days ago