What is ML.NET?

Load data

Machine learning uses known data (for example, training data) to find patterns in order to make predictions on new, unknown data.

The inputs for machine learning are called Features, which are the attributes used to make predictions. The output of machine learning is called the Label, which is the actual prediction.

Data in ML.NET is represented as an IDataView, which is a flexible, efficient way of describing tabular data (for example, rows and columns). IDataView objects can contain numbers, text, booleans, vectors, and more. You can load data from files or from real-time streaming sources to an IDataView.

LoadFromTextFile allows you to load data from TXT, CSV, TSV, and other file formats.

IDataView trainingData = mlContext.Data
    .LoadFromTextFile<SentimentInput>(dataPath, separatorChar: ',', hasHeader: true);

LoadFromEnumerable enables loading from in-memory collections, JSON/XML, relational and non-relational databases (for example, SQL, CosmosDB, MongoDB), and many other data sources.

IDataView trainingData = mlContext.Data
    .LoadFromEnumerable<SentimentInput>(inMemoryCollection);

Learn more about loading data

Transform data

In most cases, the data that you have available isn't suitable to be used directly to train a machine learning model. The raw data needs to be pre-processed using data transformations.

Transformers take data, do some work on it, and return new, transformed data.

There are built-in set of data transforms for replacing missing values, data conversion, featurizing text, and more.

// Convert sentiment text into numeric features
IEstimator<ITransformer> dataTransformPipeline = mlContext.Transforms.Text
    .FeaturizeText("Features", "SentimentText");

Explore data transforms

Choose algorithm

When using machine learning and ML.NET, you must choose a machine learning task that goes along with your scenario. ML.NET offers over 30 algorithms (or trainers) for a variety of ML tasks:

IEstimator<ITransformer> trainer = mlContext.BinaryClassification.Trainers
    .AveragedPerceptron(labelColumnName: "Sentiment", featureColumnName: "Features"));

IEstimator<ITransformer> trainingPipeline = dataTransformPipeline.Append(trainer);

Explore algorithms

Train model

The data transformations and algorithms you have specified are not executed until you call the Fit() method (because of ML.NET's lazy loading approach). This is when model training happens.

An estimator takes in data, learns from the data, and creates a transformer. In the case of model training, the training data is the input, and the trained model is the output; the trained model is thus a transformer that turns the input Features from new data into output predictions.

ITransformer model = pipeline.Fit(trainingData);

Learn more about model training

Evaluate model

ML.NET offers evaluators that assess the performance of your model on a variety of metrics:

• Accuracy
• Area under the curve (AUC)
• R-Squared
• Root Mean Squared Error (RMSE)

// Make predictions on test data
IDataView predictions = model.Transform(testDataView);

// Evaluate model and return metrics
var metrics = mlContext.BinaryClassification
    .Evaluate(predictions, labelColumnName: "Sentiment");

// Print out accuracy metric
Console.WriteLine("Accuracy" + metrics.Accuracy);

Explore evaluation metrics

Deploy & consume model

You can save your trained model as a binary file that is then integrated into your .NET applications.

mlContext.Model.Save(model, trainingData, "model.zip");

Once you have saved the trained model, you can load the model in your other .NET applications.

MLContext mlContext = new MLContext();
DataViewSchema predictionPipelineSchema;
ITransformer trainedModel = mlContext.Model.Load("model.zip", out predictionPipelineSchema);

You can then use the loaded model to start making predictions. You can use the Prediction Engine, which is a convenience API used for making single predictions, or the Transform method, which is used for making batch predictions.

var predEngine = mlContext.Model.CreatePredictionEngine(model);
SentimentInput sampleComment = new SentimentInput{ SentimentText = "This is very rude!" };
SentimentOutput result = predEngine.Predict(sampleComment);
Console.WriteLine(result.Prediction);

Learn how use a trained ML model