LinearRegressionActor

The LinearRegression is a Coral Actor which calculates the outcome variable based on a linear combination of input variables. For example, it can calculate the outcome variable if the formula is as follows:

Y = a + bX + cY + dZ

X, Y and Z are provided in the trigger JSON, while a, b, c and d are coefficients which are given in the constructor of this actor.

Creating a LinearRegressionActor

The LinearRegressionActor has "type": "linearregression". The params value contains the following fields:

field	type	required	description
`weights`	JObject	yes	An object containing the feature names and coefficients of the model.
`intercept`	Double	yes	The intercept of the linear regression model (a).
`outcome`	string	no	The name of the outcome variable to use. If not provided, “score” is used.

Example

{
  "type": "linearregression",
  "params": {
    "outcome": "creditscore",
    "intercept": 3.972,
    "weights": {
      "salary": 0.47353,
      "tnxcount1month": 1.86766,
      "age": 4.52352
    }
  }
}

In this model, the formula becomes as follows:

creditscore = 3.972 + 0.47353 * salary + 1.86766 * tnxcount1month + 4.52352 * age

Trigger

The LinearRegressionActor accepts JSON objects that contain the fields that were defined in the constructor. An example of a JSON object that comes in is as follows:

{
  "salary": 3000,
  "tnxcount1month": 25,
  "age": 30
}

Based on this input object, the LinearRegressionActor would calculate the following:

The score that is calculated is

creditscore = 3.972 + (0.47353 * 3000) + (1.86766 * 25) + (4.52352 * 30) = 1606.9591.

Emit

The LinearRegressionActor enriches the received trigger JSON with the predicted score and emits it. It looks like the following:

{
   "salary": 3000,
   "tnxcount1month": 25,
   "age": 30,
   "creditscore": 1606.9591
}

State

The LinearRegressionActor does not keep any state.

Collect

The LinearRegressionActor does not collect state from other actors.

Timer

The LinearRegressionActor does not provide timer actions.