Class OPLS

constructor

• new OPLS(
      data: any[],
      labels: any[],
      options?: {
          center?: boolean;
          cvFolds?: any[];
          maxComponents?: number;
          nbFolds?: number;
          scale?: boolean;
      },
  ): OPLS

  Creates a new OPLS model from features and labels.

  Parameters

  • data: any[]

    matrix containing data (X).

  • labels: any[]

    1D Array containing metadata (Y). Numeric labels trigger regression, string labels trigger discriminant analysis.

  • Optionaloptions: {
        center?: boolean;
        cvFolds?: any[];
        maxComponents?: number;
        nbFolds?: number;
        scale?: boolean;
    } = {}

    constructor options.

    • Optionalcenter?: boolean

      should the data be centered (subtract the mean).

    • OptionalcvFolds?: any[]

      Allows to provide folds as array of objects with the arrays trainIndex and testIndex as properties.

    • OptionalmaxComponents?: number

      upper bound on the number of orthogonal components. The fit normally stops earlier (when adding a component no longer improves the cross-validated metric by at least 0.05); this is a hard cap that guarantees termination even when the metric never plateaus.

    • OptionalnbFolds?: number

      Allows to generate the defined number of folds with the training and test set chosen randomly from the data set.

    • Optionalscale?: boolean

      should the data be scaled (divide by the standard deviation).

  Returns OPLS

center

labels

means

meansY

mode

model

orthogonalScoresCV

output

predictiveScoresCV

scale

stdevs

stdevsY

tssx

tssy

yHatScoresCV

getLogs

• getLogs(): object

  get access to all the computed elements Mainly for debug and testing

  Returns object

  output object

getScores

• getScores(): { scoresX: number[][]; scoresY: number[][] }

  Returns the cross-validated predictive and orthogonal scores.

  Returns { scoresX: number[][]; scoresY: number[][] }

  the predictive (scoresX) and orthogonal (scoresY) cross-validated scores.

predict

• predict(
      features: Matrix,
      options?: { center?: boolean; scale?: boolean; trueLabels?: any[] },
  ): object

  Predict scores for new data

  Parameters

  • features: Matrix

    a matrix containing new data

  • Optionaloptions: { center?: boolean; scale?: boolean; trueLabels?: any[] } = {}

    prediction options.

    • Optionalcenter?: boolean

      should the data be centered before prediction.

    • Optionalscale?: boolean

      should the data be scaled before prediction.

    • OptionaltrueLabels?: any[]

      an array with true values to compute confusion matrix.

  Returns object

  • predictions

predictCategory

• predictCategory(
      features: Matrix,
      options?: { center?: boolean; scale?: boolean; trueLabels?: any[] },
  ): any[]

  Predicts the class of each row of new data (discriminant analysis mode).

  Parameters

  • features: Matrix

    a matrix containing new data.

  • Optionaloptions: { center?: boolean; scale?: boolean; trueLabels?: any[] } = {}

    prediction options.

    • Optionalcenter?: boolean

      should the data be centered before prediction.

    • Optionalscale?: boolean

      should the data be scaled before prediction.

    • OptionaltrueLabels?: any[]

      an array with true values to compute confusion matrix.

  Returns any[]

  the predicted class name for each row of features.

toJSON

• toJSON(): object

  Export the current model to a JSON object

  Returns object

  model

Staticload

• load(model: object): OPLS

  Load an OPLS model from JSON

  Parameters

  • model: object

    the serialized model to load.

  Returns OPLS

  the loaded OPLS model.