8.3.3 - Using PredictionThing to perform real time scoring from a Service (not a REST call)

[SajidPatel]

Has anyone been able to do a real time scoring in 8.3 by calling 'RealtimeScore' service of 'AnalyticsServer_PredictionThing' Thing? I want to be able to send data constructed as an infotable. 

I have a published Analytics Model. I tried the following code without much luck:

var adsmf = Resources["InfoTableFunctions"].CreateInfoTableFromDataShape({
  infoTableName : "InfoTable",
  dataShapeName : "AnalyticsDatasetMetadataFlattened"
});

var dataShape = DataShapes["AnalyticsServerConnector.pa_temp_pa_lft_temp_705999f9-e266-4771-91b2-b8d38bf1ca9f.InputParamsDataShape"].GetFieldDefinitions();

var tableLength = dataShape.rows.length;
for (var x=0; x < tableLength; x++) {
  var row = dataShape.rows[x];
  
  if (row.name.startsWith('_')) {
    fieldName = newEntry.item.substring(1, newEntry.item.length);

    var adsmfEntry = new Object();
    adsmfEntry.timeSamplingInterval = undefined; // INTEGER
    adsmfEntry.isStatic = undefined; // BOOLEAN
    adsmfEntry.fieldName = fieldName; // STRING
    adsmfEntry.min = row.min; // NUMBER
    adsmfEntry.max = row.max; // NUMBER
    adsmfEntry.dataType = row.dataType; // STRING
    adsmfEntry.opType = row.opType; // STRING
  }
}

var dsRef = Resources["InfoTableFunctions"].CreateInfoTableFromDataShape({
  infoTableName : "InfoTable",
  dataShapeName : "AnalyticsDatasetRef"
});

// Tried different way to set values without any luck - any clues would be helpful

dsRefEntry = new Object();
dsRefEntry.datasetUri = 'body:/'; // STRING
dsRefEntry.metadata = adsmf; // INFOTABLE
//dsRefEntry.data = values; // INFOTABLE
dsRefEntry.format = 'csv'; // STRING
dsRef.AddRow(dsRefEntry);



var result = Things["PTC02PWIL-AnalyticsServer_PredictionThing"].RealtimeScore({
  goalField: 'MY_GOAL' /* STRING */,
  modelUri: 'http://localhost:8080/Thingworx/FileRepositories/TW.AnalysisServices.ModelRepository/AnalyticsServerConnector/pa_temp_pa_lft_temp_705999f9-e266-4771-91b2-b8d38bf1ca9f.xml',
  identifierFields: idFields,
  datasetRef: dsRef
});

[cmorfin · Accepted solution]

Hi @SajidPatel 

We have a couple of post that do use , among other things, the RealTimeScore service of PredictionThing.

They come with the entities that you can import and review the code.

Analytics: Prediction Methods Mashup

How to work with ordinal and categorical data in ThingWorx Analytics 

Also in your code the modelURI has a wrong format, it should be of type results://models/<jobID>

Similarly you will be able to see this in the above posts.

Finally be mindful of https://www.ptc.com/en/support/article?n=CS298126 if passing scoring data inline using a body type, it will only work with small amount of data.

How this helps

Kind regards

Christophe

[SajidPatel]

@cmorfin Thanks, this is exactly what I was looking for. I am scoring just one row of data on need basis.