ML

The

ML module facilitates creation of Automated Machine Learning models to embed into Templates.

Example

  // use a DictType stream
  const training_stream = Stream(
      "My Training Stream",
      DictType(
          StringType,
          StructType({
              qty: FloatType,
              price: FloatType,
          })
      )
  );

  const prediction_stream = Stream(
      "My Prediction Stream",
      DictType(
          StringType,
          StructType({
              price: FloatType,
          })
      )
  );

  // create a GP to predict demand based on price
  const demand = new MLModelBuilder("Demand")
      .feature("price", FloatType)
      .output(FloatType)
      .trainFromStream({
          output_name: "qty",
          input: training_stream,
      });

  // directly predict the model outputs from features
  const ml_prediction = new MLPredictionBuilder("ML Prediction")
    .model(demand)
    .predict(prediction_stream);

  // or alternatively use the model to predict in a process
  // create the sales process and add the ml function
  // which is evaluated in each sale
  const sales = new ProcessBuilder("sales")
      // the ml function can be added
      .ml(demand)
      // the sale needs a price
      .value("price", FloatType)
      // the ml can be evaluated
      .let(
          "qty",
          (props, _resources, mls) => mls.demand(Struct({ price: props.price }))
      )
      .mapFromValue({ date: new Date(0), price: 10 });

Classes

• FloatMLBuilder
• MLExamplesEvaluatorBuilder
• MLModelBuilder
• MLModelEvaluatorBuilder
• MLPredictionBuilder

Other

FloatMLModel

Ƭ FloatMLModel: Object

The machine learning model applicable to

FloatType outputs

Type declaration

Name	Type
noise	FloatMLModelNoise
type	FloatMLModelType

---

FloatMLModelNoise

Ƭ FloatMLModelNoise: "none" | "gaussian"

The noise modelled in the

FloatMLModel output.

---

FloatMLModelType

Ƭ FloatMLModelType: "boosted_trees" | "constant" | "gaussian_process" | "linear" | "neural_network"

The machine learning model types that can be used for a

FloatMLModel.

---

MLModelType

Ƭ MLModelType:

FloatMLModelType | StringMLModelType

The machine learning model types that can be used for a MLModel.

---

MLPredictionConfiguration

Ƭ MLPredictionConfiguration<T>: T extends

FloatType ? { max: EastFunction<FloatType> ; min: EastFunction<FloatType> } : Record<string, never>

Extra configuration that can modify ML evaluation

Type parameters

Name	Type
T	extends EastType = EastType

---

StringMLModel

Ƭ StringMLModel: Object

The machine learning model applicable to

StringType outputs

Type declaration

Name	Type
noise	StringMLModelNoise
type	StringMLModelType

---

StringMLModelNoise

Ƭ StringMLModelNoise: "maximum_likelihood" | "weighted"

The noise modelled in the

StringMLModel output.

---

StringMLModelType

Ƭ StringMLModelType: "boosted_trees_string" | "neural_network_string"

The machine learning model types that can be used for a

StringMLModel.