機械学習パイプライン

エンドツーエンドの機械学習ワークフローをパイプラインとして実行および展開する

パイプラインは、データ処理ワークフローの複数のステップを結合して整理したものです。機械学習の場合、これらのステップには、データ準備、特徴量エンジニアリング、特徴選択、モデル化、後処理が含まれます。パイプラインを実行すると、パイプラインを通過するときに各ステップがデータに適用されます。データ前処理専用や特徴量エンジニアリング専用のパイプラインを構築できます。あるいは、データ前処理、特徴量エンジニアリング、分類または回帰、および推論の複数のステップを一緒にした機械学習パイプラインを作成することもできます。

オブジェクト

すべて展開する

パイプラインオブジェクト

LearningPipeline Machine learning pipeline (R2026a 以降)

データ前処理のコンポーネント

`equalWidthBinnerComponent`	Pipeline component for grouping data into equal-width bins (R2026a 以降)
`frequencyEncoderComponent`	Pipeline component for frequency encoding categorical variables (R2026a 以降)
`kmeansEncoderComponent`	Pipeline component for feature extraction using k-means clustering (R2026a 以降)
`normalizerComponent`	Pipeline component for normalizing data (R2026a 以降)
`observationImputerComponent`	Pipeline component for imputing missing values (R2026a 以降)
`observationRemoverComponent`	Pipeline component for removing observations (R2026a 以降)
`oneHotEncoderComponent`	Pipeline component for encoding categorical data into one-hot vectors (R2026a 以降)
`outlierImputerComponent`	Pipeline component for imputing outlier values (R2026a 以降)
`outlierRemoverComponent`	Pipeline component for removing outlier values (R2026a 以降)
`pcaComponent`	Pipeline component for principal component analysis (PCA) (R2026a 以降)
`quantileBinnerComponent`	Pipeline component for binning data based on quantiles (R2026a 以降)
`ricaComponent`	Pipeline component for feature extraction using reconstruction independent component analysis (RICA) (R2026a 以降)
`sparseFilterComponent`	Pipeline component for feature extraction using sparse filtering (R2026a 以降)

特徴選択および特徴量エンジニアリングのコンポーネント

`featureSelectionClassificationANOVAComponent`	Pipeline component for performing feature selection using ANOVA algorithm (R2026a 以降)
`featureSelectionClassificationChi2Component`	Pipeline component for performing feature selection using chi-square tests (R2026a 以降)
`featureSelectionClassificationKruskalWallisComponent`	Pipeline component for performing feature selection using Kruskal-Wallis test (R2026a 以降)
`featureSelectionClassificationMRMRComponent`	Pipeline component for performing MRMR feature selection in classification workflow (R2026a 以降)
`featureSelectionClassificationNCAComponent`	Pipeline component for performing feature selection using neighborhood component analysis (NCA) for classification (R2026a 以降)
`featureSelectionClassificationReliefFComponent`	Pipeline component for performing feature selection using ReliefF algorithm (R2026a 以降)
`featureSelectionRegressionFTestComponent`	Pipeline component for performing feature selection using F-tests (R2026a 以降)
`featureSelectionRegressionMRMRComponent`	Pipeline component for performing MRMR feature selection in regression workflow (R2026a 以降)
`featureSelectionRegressionNCAComponent`	Pipeline component for performing feature selection using neighborhood component analysis (NCA) for regression (R2026a 以降)
`featureSelectionRegressionReliefFComponent`	Pipeline component for performing feature selection using RReliefF algorithm (R2026a 以降)
`variableSelectorComponent`	Pipeline component for manual variable selection (R2026a 以降)

教師あり学習のコンポーネント

分類のコンポーネント

`classificationDiscriminantComponent`	Pipeline component for discriminant analysis classification (R2026a 以降)
`classificationECOCComponent`	Pipeline component for multiclass classification using error-correcting output codes (ECOC) model (R2026a 以降)
`classificationEnsembleComponent`	Pipeline component for ensemble classification (R2026a 以降)
`classificationGAMComponent`	Pipeline component for binary classification using generalized additive model (GAM) (R2026a 以降)
`classificationKernelComponent`	Pipeline component for classification using Gaussian kernel with random feature expansion (R2026a 以降)
`classificationKNNComponent`	Pipeline component for classification using k-nearest neighbor model (R2026a 以降)
`classificationLinearComponent`	Pipeline component for binary classification of high-dimensional data using linear model (R2026a 以降)
`classificationNaiveBayesComponent`	Pipeline component for multiclass classification using naive Bayes model (R2026a 以降)
`classificationNeuralNetworkComponent`	Pipeline component for classification using neural network model (R2026a 以降)
`classificationSVMComponent`	Pipeline component for one-class and binary classification using SVM classifier (R2026a 以降)
`classificationTreeComponent`	Pipeline component for multiclass classification using binary decision trees (R2026a 以降)

回帰のコンポーネント

`regressionEnsembleComponent`	Pipeline component for regression using ensemble of learners (R2026a 以降)
`regressionGAMComponent`	Pipeline component for generalized additive model (GAM) for regression (R2026a 以降)
`regressionGPComponent`	Pipeline component for Gaussian process regression (GPR) (R2026a 以降)
`regressionLinearComponent`	Pipeline component for regression of high-dimensional data using a linear model (R2026a 以降)
`regressionKernelComponent`	Pipeline component for regression using Gaussian kernel model (R2026a 以降)
`regressionNeuralNetworkComponent`	Pipeline component for regression using neural network model (R2026a 以降)
`regressionSVMComponent`	Pipeline component for regression using a support vector machine (SVM) model (R2026a 以降)
`regressionTreeComponent`	Pipeline component for regression using binary decision trees (R2026a 以降)

カスタムコンポーネント

functionComponent Pipeline component for custom function (R2026a 以降)

関数

すべて展開する

作成と変更

自動接続

`series`	Connect components in series to create pipeline (R2026a 以降)
`parallel`	Connect components or pipelines in parallel to create pipeline (R2026a 以降)
`insert`	Insert component or pipeline into existing pipeline (R2026a 以降)
`replace`	Replace existing pipeline component with new component (R2026a 以降)

手動接続

`add`	Add new component or pipeline to existing pipeline (R2026a 以降)
`remove`	Remove existing components or pipelines from pipeline (R2026a 以降)
`connect`	Create connections between pipeline components (R2026a 以降)
`disconnect`	Remove connections between ports in pipeline (R2026a 以降)

階層

expand Expand subpipelines in pipeline (R2026a 以降)

実行

`learn`	Initialize and evaluate pipeline or component (R2026a 以降)
`run`	Execute pipeline or component for inference after learning (R2026a 以降)
`prune`	Remove unnecessary components and dependencies from pipeline (R2026a 以降)
`reset`	Reset pipeline or component (R2026a 以降)
`crossvalidate`	Cross-validate pipeline (R2026a 以降)

可視化と記述

`view`	View diagram of pipeline inputs, outputs, components, and connections (R2026a 以降)
`describe`	Display summary of pipeline components (R2026a 以降)