tsc_classifiers: Return available classifiers

In compstat-lmu/TSClassification: Interface to various time series classifiers from the "The Great Time Series Classification Bake Off"

Description

Run tsc_classifiers() to obtain available classifiers

Usage

tsc_classifiers()

Details

The following classifiers are available:

Value

character Names of available classifiers.

Ensemble Classifiers

• timeseriesweka.classifiers.ElasticEnsemble
  Combination of nearest Neighbour (NN) classifiers that use elastic distance measures
  Hyperparameters: None

• timeseriesweka.classifiers.FlatCote
  Collective of Transformation Ensembles (Bagnall et al.,2015)
  Hyperparameters: None

  Base-learners of ElasticEnsemble:
  

• timeseriesweka.classifiers.ensembles.elastic_ensemble.WDTW1NN
  Elastic Ensemble of Nearest Neighbour Algorithms:
  Weighted Dynamic Time Warping 1 Nearest Neighbour
  Hyperparameters: None

• timeseriesweka.classifiers.ensembles.elastic_ensemble.ED1NN
  Euclidean distance with 1 nearest neighbor
  Hyperparameters: None

• timeseriesweka.classifiers.ensembles.elastic_ensemble.DTW1NN
  Dynamic time warping with 1 nearest neighbor
  Hyperparameters:
  

  • setWindow: double range: [1, Inf]

• timeseriesweka.classifiers.ensembles.elastic_ensemble.ERP1NN
  edit distance with real penalty with 1 nearest neighbor
  Hyperparameters: None

• timeseriesweka.classifiers.ensembles.elastic_ensemble.LCSS1NN
  longest common subsequence with 1 nearest neighbor
  Hyperparameters: None

• timeseriesweka.classifiers.ensembles.elastic_ensemble.TWE1NN
  Time Warp Edit with 1 nearest neighbor
  Hyperparameters: None

• timeseriesweka.classifiers.ensembles.elastic_ensemble.MSM1NN
  Move-Split-Merge with 1 nearest neighbor
  Hyperparameters: None

Differential Distance Based Classifiers

• timeseriesweka.classifiers.NN_CID
  Complexity Invariant distance with k nearest neighbor
  Hyperparameters: None

• timeseriesweka.classifiers.DD_DTW
  Derivative dynamic time warping
  Hyperparameters: None

• timeseriesweka.classifiers.DTD_C
  Derivative transform distance
  Hyperparameters: None

Dictionary based Classifiers

• timeseriesweka.classifiers.BOSS
  Bag of SFA Symbols
  Hyperparameters:
  

  • setMaxEnsembleSize: integer(1) range: [1, Inf]

• timeseriesweka.classifiers.BagOfPatterns
  Bag of Patterns
  Hyperparameters: None
  

• timeseriesweka.classifiers.SAX_1NN
  Symbolic Aggregate Approximation
  Hyperparameters: None
  

• timeseriesweka.classifiers.SAXVSM
  Symbolic Aggregate Approximation - Vector Space Model
  Hyperparameters: None
  

Shapelet based Classifiers

• timeseriesweka.classifiers.ShapeletTransformClassifier
  Shapelet transformation that separates the Shapelet discovery from the classifier by finding the top k Shapelets in a single run
  Hyperparameters:

  • setTransformType: character(1)
    values: "univariate","uni","shapeletd","shapeleti"

  • setNumberOfShapelets: integer(1) range: [1, Inf]

• timeseriesweka.classifiers.FastShapelets
  Fast Shapelets (FS)
  Hyperparameters: None
  

• timeseriesweka.classifiers.LearnShapelets
  Learned Shapelets (LS):
  Hyperparameters: None
  

Interval based Classifiers

• timeseriesweka.classifiers.TSF
  Time Series Forest (Deng et al.,2013)
  Hyperparameters:
  

  • setNumTrees: integer(1) range: [1, Inf]

• timeseriesweka.classifiers.TSBF
  Time Series Bag of Features (TSBF)
  Hyperparameters:
  

  • setZLevel: double(1)

• timeseriesweka.classifiers.LPS
  Learned Pattern Similarity (LPS)
  Hyperparameters: None
  

Time Series Classifier

• timeseriesweka.classifiers.DTW_kNN
  specialization of kNN that can only be used with the efficient DTW distance
  Hyperparameters:
  

  • setMaxR: double(1) range: \[0, 1] set max window size

• timeseriesweka.classifiers.FastDTW_1NN
  fast Dynamic time warping with 1 nearest neighbor
  Hyperparameters:
  

  • setR: double(1)

• timeseriesweka.classifiers.SlowDTW_1NN
  compare with FastDTW_1NN
  Hyperparameters:
  

  • setR: double(1)

Weka Classifiers

Several WEKA classifiers have been implemented in the Time-Series Classification Bake-off.
The use of those classifiers is discouraged from within TSClassification, but nonetheless implemented for completeness.
We advise to use the official implementations from package RWeka (https://cran.r-project.org/web/packages/RWeka/index.html) for greater flexibility and improved support for setting hyperparameters.

• weka.classifiers.functions.Logistic

• weka.classifiers.bayes.BayesNet

• weka.classifiers.bayes.NaiveBayes

• weka.classifiers.functions.Logistic

• weka.classifiers.functions.MultilayerPerceptron

• weka.classifiers.functions.SMO

• weka.classifiers.meta.RotationForest

• weka.classifiers.trees.J48

• weka.classifiers.trees.RandomForest

Examples

tsc_classifiers()

compstat-lmu/TSClassification documentation built on Nov. 18, 2019, 6:56 a.m.