The Trainer class

In the Trainer class, we will build a new pipeline to train our model. The FeaturizeText transform builds NGrams from the strings data we previously extracted from the files. NGrams are a popular method to create vectors from a string to, in turn, feed the model. You can think of NGrams as breaking a longer string into ranges of characters based on the value of the NGram parameter. A bi-gram, for instance, would take the following sentence, ML.NET is great and convert it into ML-.N-ET-is-gr-ea-t. Lastly, we build the SdcaLogisticRegression trainer object:

var dataProcessPipeline = MlContext.Transforms.CopyColumns("Label", nameof(FileInput.Label))
 .Append(MlContext.Transforms.Text.FeaturizeText("NGrams", nameof(FileInput.Strings)))
 .Append(MlContext.Transforms.Concatenate("Features", "NGrams"));

var trainer = MlContext.BinaryClassification.Trainers.SdcaLogisticRegression(labelColumnName: "Label", featureColumnName: "Features");