[SOLVED] Receiving, "An error was thrown and was not caught: The validation data provided must contain ..." when creating a Text Classifier Model with CreateML

Receiving, "An error was thrown and was not caught: The validation data provided must contain ..." when creating a Text Classifier Model with CreateML

I am using Playground to create a Text Classifier Model using CreateML and keep getting the error:

Playground execution terminated: An error was thrown and was not caught:
▿ The validation data provided must contain class.
  ▿ type : 1 element
    - reason : "The validation data provided must contain class."

My code is relatively simple, using two columns from a data table. The textColumn is labeled "text" and the labelColumn is labeled "class":

import Cocoa
import CreateML

let data = try MLDataTable(contentsOf: URL(fileURLWithPath: "/Users/ ... .csv"))
let(trainingData, testingData) = data.randomSplit(by: 0.8, seed: 5)
let sentimentClassifier = try MLTextClassifier(trainingData: trainingData, textColumn: "text", labelColumn: "class")
let evaluationMetrics = sentimentClassifier.evaluation(on: testingData, textColumn: "text", labelColumn: "class")
let evaluationAccuracy = (1.0 - evaluationMetrics.classificationError) * 100

The only difference I can find between this and the code provided in the Apple Developer Documentation is that instead of

let evaluationMetrics = sentimentClassifier.evaluation(on: testingData, textColumn: "text", labelColumn: "class")

their documentation is:

let evaluationMetrics = sentimentClassifier.evaluation(on: testingData)

and version 11.2.1 of Xcode gives me a failure if I try using the line from the Apple Developer Documentation.

Thanks in advance for any help you can offer.

Solution

Try This! it works for me

let data = try MLDataTable(contentsOf: URL(fileURLWithPath: "/Users/justinmacbook/Desktop/twitter-sanders-apple3.csv"))

let (trainingData, testingData) = data.randomSplit(by: 0.8, seed: 5 )

let sentimentClassifier = try MLTextClassifier(trainingData: trainingData, textColumn: "class", labelColumn: "text")

let evaluationMetrics = sentimentClassifier.evaluation(on: testingData, textColumn: "class", labelColumn: "text")

let evaluationAccuracy = (1.0 - evaluationMetrics.classificationError) * 100