使用 CreateML 创建文本分类器模型时收到“引发错误但未捕获：提供的验证数据必须包含...”

Posted 2023-03-28

【中文标题】使用 CreateML 创建文本分类器模型时收到“引发错误但未捕获：提供的验证数据必须包含...”

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

【发布时间】：2020-04-25 11:16:03

【问题描述】：

我正在使用 Playground 使用 CreateML 创建文本分类器模型并不断收到错误：

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."

我的代码比较简单，使用数据表中的两列。 textColumn 标记为“text”，labelColumn 标记为“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

我能发现这与 Apple 开发者文档中提供的代码之间的唯一区别是，而不是

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

他们的文档是：

let evaluationMetrics = sentimentClassifier.evaluation(on: testingData)

如果我尝试使用 Apple Developer Documentation 中的行，Xcode 的 11.2.1 版会失败。

提前感谢您提供的任何帮助。

【问题讨论】：

你找到解决办法了吗？

【参考方案1】：

试试这个！它对我有用

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

【讨论】：

请正确格式化代码，以便其他人（包括未来用户）实际上可读且易于理解

那行得通@PrinceRana。唯一的问题是我交换了 textColumn 和 labelColumn 标签。谢谢。

【参考方案2】：

这是对我有用的解决方案。我相信这个原始代码只能在 OSX 10.15+ 上运行

import Cocoa
import CreateML
import NaturalLanguage

let data = try MLDataTable(contentsOf: URL(fileURLWithPath: "/Users/m0a04y6/Desktop/ios/ML/twitter-sanders-apple3.csv"))

let (trainingData, testingData) = data.randomSplit(by: 0.8, seed: 5)
    
let parameters = MLTextClassifier.ModelParameters.init(validationData: trainingData, algorithm: MLTextClassifier.ModelAlgorithmType.maxEnt(revision: 1), language: NLLanguage.english, textColumnValidationData: "text", labelColumnValidationData: "class")
let sentimentClassifier = try MLTextClassifier(trainingData: trainingData, textColumn: "text", labelColumn: "class", parameters: parameters)

let evaluationMetrics = sentimentClassifier.evaluation(on: testingData, textColumn: "text", labelColumn: "class")     //Training accuracy as a percentage

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

let metadata = MLModelMetadata(author: "Madhur Ahuja", shortDescription: "A model trained to classify movie review sentiment", version: "1.0")
try sentimentClassifier.write(to: URL(fileURLWithPath: "/Users/m0a04y6/Desktop/iOS/ML/sentiment.mlmodel"), metadata: metadata)