DataPredict

API Reference - Models - AffinityPropagation

AffinityPropagation is an unsupervised machine learning model that predicts which cluster that the input belongs to using a similarity matrix that measures the similarity between each data point.

Stored Model Parameters

Contains a table.

• ModelParameters[1]: Contains previously stored feature matrix.

• ModelParameters[2]: Contains the cluster number array that has a length of m, where m is the number of data from previous feature matrix and its values contains the cluster numbers indicating which of the clusters that each feature vector belongs to.

• ModelParameters[3]: Contains the responsibility matrix of (m x m) size, where m is the number of data.

• ModelParameters[4]: Contains the availability matrix of (m x m) size, where m is the number of data.

Constructors

new()

Create new model object. If any of the arguments are nil, default argument values for that argument will be used.

AffinityPropagation.new(maximumNumberOfIterations: integer, distanceFunction: string, preferenceType: string, damping: number, preferenceValueArray: {number}): ModelObject

Parameters:

• maximumNumberOfIterations: How many times should the model needed to be trained.

• distanceFunction: The distance function to be used. Available options are:

  • Euclidean (Default)

  • Manhattan

  • Cosine

• preferenceType: Determines how preferences are calculated. Available options are:

  • Median (Default)

  • Average

  • Minimum

  • Maximum

  • Precomputed

• damping: A high value leads to fewer changes, while a low value leads to more exploration. The value can be set between 0 and 1.

• preferenceValueArray: An array containing preference values. The index determines the preference value for number of data for that index. Can only be used with “Precomputed” preferenceType.

Returns:

• ModelObject: The generated model object.

Functions

setParameters()

Set model’s parameters. When any of the arguments are nil, previous argument values for that argument will be used.

AffinityPropagation:setParameters(maximumNumberOfIterations: integer, distanceFunction: string, preferenceType: string, damping: number, preferenceValueArray: {number})

Parameters:

• maximumNumberOfIterations: How many times should the model needed to be trained.

• distanceFunction: The distance function to be used. Available options are:

  • Euclidean

  • Manhattan

  • Cosine

• preferenceType: Determines how preferences are calculated. Available options are:

  • Median

  • Average

  • Minimum

  • Maximum

  • Precomputed

• damping: A high value leads to fewer changes, while a low value leads to more exploration. The value can be set between 0 and 1.

• preferenceValueArray: An array containing preference values. The index determines the preference value for number of data for that index. Can only be used with “Precomputed” preferenceType.

train()

Train the model.

AffinityPropagation:train(featureMatrix: Matrix)

Parameters:

• featureMatrix: Matrix containing all data.

Returns:

• costArray: An array containing cost values.

predict()

Predict which cluster does it belong to for a given data.

AffinityPropagation:predict(featureMatrix: Matrix): Matrix, Matrix

Parameters:

• featureMatrix: Matrix containing data.

Returns:

• clusterNumberVector: The cluster which the data belongs to.

• shortestDistanceVector: The distance between the datapoint and the center of the cluster (centroids).

Inherited From

• BaseModel