cluster

Documentation

Index

• func ComputeWeights(X *DenseMatrix, imp float64) []float64
• func CreateFrequencyTable(X *DenseMatrix) [][]KV
• func EuclideanDistance(a, b *DenseVector) (float64, error)
• func HammingDistance(a, b *DenseVector) (float64, error)
• func SetWeights(newWeights []float64)
• func WeightedHammingDistance(a, b *DenseVector) (float64, error)
• type DenseMatrix
  • func InitCao(X *DenseMatrix, clustersNumber int, distFunc DistanceFunction) (*DenseMatrix, error)
  • func InitHuang(X *DenseMatrix, clustersNumber int, distFunc DistanceFunction) (*DenseMatrix, error)
  • func InitNum(X *DenseMatrix, clustersNumber int, distFunc DistanceFunction) (*DenseMatrix, error)
  • func InitRandom(X *DenseMatrix, clustersNumber int, distFunc DistanceFunction) (*DenseMatrix, error)
  • func NewDenseMatrix(r, c int, data []float64) *DenseMatrix
• type DenseVector
  • func NewDenseVector(n int, data []float64) *DenseVector
• type DistanceFunction
• type InitializationFunction
• type KModes
  • func NewKModes(dist DistanceFunction, init InitializationFunction, clusters int, runs int, ...) *KModes
  • func (km *KModes) FitModel(X *DenseMatrix) error
  • func (km *KModes) LoadModel() error
  • func (km *KModes) Predict(X *DenseMatrix) (*DenseVector, error)
  • func (km *KModes) SaveModel() error
• type KPrototypes
  • func NewKPrototypes(dist DistanceFunction, init InitializationFunction, categorical []int, ...) *KPrototypes
  • func (km *KPrototypes) FitModel(X *DenseMatrix) error
  • func (km *KPrototypes) LoadModel() error
  • func (km *KPrototypes) Predict(X *DenseMatrix) (*DenseVector, error)
  • func (km *KPrototypes) SaveModel() error
• type KV

Functions

func ComputeWeights

func ComputeWeights(X *DenseMatrix, imp float64) []float64

ComputeWeights derives weights based on the frequency of attribute values (more different values means lower weight).

func CreateFrequencyTable

func CreateFrequencyTable(X *DenseMatrix) [][]KV

CreateFrequencyTable creates frequency table for attributes in given matrix, it returns attributes in frequency descending order.

func EuclideanDistance

func EuclideanDistance(a, b *DenseVector) (float64, error)

EuclideanDistance computes eucdlidean distance between two vectors.

func HammingDistance

func HammingDistance(a, b *DenseVector) (float64, error)

HammingDistance is a basic dissimilarity function for the kmodes algorithm.

func SetWeights

func SetWeights(newWeights []float64)

SetWeights sets the weight vector used in WeightedHammingDistance function.

func WeightedHammingDistance

func WeightedHammingDistance(a, b *DenseVector) (float64, error)

WeightedHammingDistance dissimilarity function is based on hamming distance but it adds improttance to attributes.

Types

type DenseMatrix

type DenseMatrix struct {
	*mat.Dense
}

DenseMatrix wraps *gonum.org/v1/gonum/mat.Dense type.

func InitCao

func InitCao(X *DenseMatrix, clustersNumber int, distFunc DistanceFunction) (*DenseMatrix, error)

InitCao implements initialization of cluster centroids based on the frequency and density of attributes as defined in

"A new initialization method for categorical data clustering" by F.Cao(2009)

func InitHuang

func InitHuang(X *DenseMatrix, clustersNumber int, distFunc DistanceFunction) (*DenseMatrix, error)

InitHuang implements initialization of cluster centroids based on the frequency of attributes as defined in paper written by Z.Huang in 1998.

func InitNum

func InitNum(X *DenseMatrix, clustersNumber int, distFunc DistanceFunction) (*DenseMatrix, error)

InitNum initializes cluster centers for numerical data - random initialization.

func InitRandom

func InitRandom(X *DenseMatrix, clustersNumber int, distFunc DistanceFunction) (*DenseMatrix, error)

InitRandom randomly initializes cluster centers - vectors chosen from X table.

func NewDenseMatrix

func NewDenseMatrix(r, c int, data []float64) *DenseMatrix

NewDenseMatrix creates new DenseMatrix

type DenseVector

type DenseVector struct {
	*mat.VecDense
}

DenseVector wraps *gonum.org/v1/gonum/mat.VecDense type.

func NewDenseVector

func NewDenseVector(n int, data []float64) *DenseVector

NewDenseVector creates new DenseVector

type DistanceFunction

type DistanceFunction func(a, b *DenseVector) (float64, error)

DistanceFunction compute distance between two vectors.

type InitializationFunction

type InitializationFunction func(X *DenseMatrix, clustersNumber int, distFunc DistanceFunction) (*DenseMatrix, error)

InitializationFunction compute initial vales for cluster_centroids_.

type KModes

type KModes struct {
	DistanceFunc       DistanceFunction
	InitializationFunc InitializationFunction
	ClustersNumber     int
	RunsNumber         int
	MaxIterationNumber int
	WeightVectors      [][]float64
	FrequencyTable     [][]map[float64]float64 // frequency table - list of lists with dictionaries containing frequencies of values per cluster and attribute
	LabelsCounter      []int
	Labels             *DenseVector
	ClusterCentroids   *DenseMatrix
	IsFitted           bool
	ModelPath          string
}

KModes is a basic class for the k-modes algorithm, it contains all necessary information as alg. parameters, labels, centroids, ...

func NewKModes

func NewKModes(dist DistanceFunction, init InitializationFunction, clusters int, runs int, iters int, weights [][]float64, modelPath string) *KModes

NewKModes implements constructor for the KModes struct.

func (*KModes) FitModel

func (km *KModes) FitModel(X *DenseMatrix) error

FitModel main algorithm function which finds the best clusters centers for the given dataset X. func (km *KModes) FitModel(X *mat.Dense) error {

func (*KModes) LoadModel

func (km *KModes) LoadModel() error

LoadModel loads model (KModes struct) from file.

func (*KModes) Predict

func (km *KModes) Predict(X *DenseMatrix) (*DenseVector, error)

Predict assign labels for the set of new vectors.

func (*KModes) SaveModel

func (km *KModes) SaveModel() error

SaveModel saves computed ml model (KModes struct) in file specified in configuration.

type KPrototypes

type KPrototypes struct {
	DistanceFunc        DistanceFunction
	InitializationFunc  InitializationFunction
	CategoricalInd      []int
	ClustersNumber      int
	RunsNumber          int
	MaxIterationNumber  int
	WeightVectors       [][]float64
	FrequencyTable      [][]map[float64]float64 // frequency table - list of lists with dictionaries containing frequencies of values per cluster and attribute
	MembershipNumTable  [][]float64             // membership table for numeric attributes - list of labels for each cluster
	LabelsCounter       []int
	Labels              *DenseVector
	ClusterCentroids    *DenseMatrix
	ClusterCentroidsCat *DenseMatrix
	ClusterCentroidsNum *DenseMatrix
	Gamma               float64
	IsFitted            bool
	ModelPath           string
}

KPrototypes is a basic class for the k-prototypes algorithm, it contains all necessary information as alg. parameters, labels, centroids, ...

func NewKPrototypes

func NewKPrototypes(dist DistanceFunction, init InitializationFunction, categorical []int, clusters int, runs int, iters int, weights [][]float64, g float64, modelPath string) *KPrototypes

NewKPrototypes implements constructor for the KPrototypes struct.

func (*KPrototypes) FitModel

func (km *KPrototypes) FitModel(X *DenseMatrix) error

FitModel main algorithm function which finds the best clusters centers for the given dataset X.

func (*KPrototypes) LoadModel

func (km *KPrototypes) LoadModel() error

LoadModel loads model (KPrototypes struct) from file.

func (*KPrototypes) Predict

func (km *KPrototypes) Predict(X *DenseMatrix) (*DenseVector, error)

Predict assign labels for the set of new vectors.

func (*KPrototypes) SaveModel

func (km *KPrototypes) SaveModel() error

SaveModel saves computed ml model (KPrototypes struct) in file specified in configuration.

type KV

type KV struct {
	Key   float64
	Value float64
}

KV is a structure that holds key-value pairs of type float64.