Question: How to get accuracy in predicted label

Is it possible to get the accuracy/precision of a predicted label? Is there an attribute I can use the extract it? Basically some thing like:

public class EmployeePrediction
{
    [ColumnName("PredictedLabel")]
    public bool Outcome;

    [SomeAttribute]
    public float Accuracy;
}

Accuracy is actually a cumulative measure. It is computed on all the predictions in the dataset. You would like to see the following sample to know how to compute metrics for binary classification task.

https://github.com/dotnet/machinelearning-samples/tree/master/samples/getting-started/BinaryClassification_SentimentAnalysis

Basically, this what you would like to do after the model is trained.

var testData = new TextLoader(TestDataPath).CreateFrom<SentimentData>();
var evaluator = new BinaryClassificationEvaluator();
var metrics = evaluator.Evaluate(model, testData);

Variable metrics will contain all the computed metrics relevant to binary classification task.

Got it! Thanks.

Is there a way to bring the metric data into the model.zip file ? Can I add it to the predictionmodel in some way? The scenario is that when the prediction comes as a response from, lets say an api that uses my trained model, I want to tell the user the certain percentage that the prediction is correct.

Rev OVA pipeline node SDK output: wrap binary trainers as children in…

code migration from machinelearning-automl to machinelearning repo fe…

Rev OVA pipeline node SDK output: wrap binary trainers as children in…

Added onnx export functionality for LpNormNormalizingTransformer (#4161)

Added Onnx export functionality to PCATransformer (#4188)

Navigation Menu

Repository navigation

Question: How to get accuracy in predicted label #317

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Question: How to get accuracy in predicted label #317

Description

Activity

zeahmed commented on Jun 6, 2018

epignosisx commented on Jun 7, 2018

ridj87 commented on Jul 12, 2018

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions