Pipeline di Machine Learning

Eseguire e distribuire workflow di Machine Learning end-to-end come pipeline

Una pipeline combina e organizza le diverse fasi di un workflow di elaborazione dei dati. Per quanto riguarda il Machine Learning, queste fasi comprendono la preparazione dei dati, l'ingegneria delle feature e la loro selezione, la modellazione e la post-elaborazione. L'esecuzione di una pipeline comporta l'applicazione di ciascuna fase ai dati man mano che passano attraverso la pipeline. È possibile creare una pipeline dedicata esclusivamente alla pre-elaborazione dei dati o esclusivamente all'ingegneria delle feature. In alternativa, è possibile creare una pipeline di Machine Learning che comprenda più fasi: pre-elaborazione dei dati, ingegneria delle feature, classificazione o regressione e inferenza, tutte integrate in un unico processo.

Oggetti

espandi tutto

LearningPipelineMachine learning pipeline (Da R2026a)
equalWidthBinnerComponentPipeline component for grouping data into equal-width bins (Da R2026a)
frequencyEncoderComponentPipeline component for frequency encoding categorical variables (Da R2026a)
kmeansEncoderComponentPipeline component for feature extraction using k-means clustering (Da R2026a)
normalizerComponentPipeline component for normalizing data (Da R2026a)
observationImputerComponentPipeline component for imputing missing values (Da R2026a)
observationRemoverComponentPipeline component for removing observations (Da R2026a)
oneHotEncoderComponentPipeline component for encoding categorical data into one-hot vectors (Da R2026a)
outlierImputerComponentPipeline component for imputing outlier values (Da R2026a)
outlierRemoverComponentPipeline component for removing outlier values (Da R2026a)
pcaComponentPipeline component for principal component analysis (PCA) (Da R2026a)
quantileBinnerComponentPipeline component for binning data based on quantiles (Da R2026a)
ricaComponentPipeline component for feature extraction using reconstruction independent component analysis (RICA) (Da R2026a)
sparseFilterComponentPipeline component for feature extraction using sparse filtering (Da R2026a)
featureSelectionClassificationANOVAComponentPipeline component for performing feature selection using ANOVA algorithm (Da R2026a)
featureSelectionClassificationChi2ComponentPipeline component for performing feature selection using chi-square tests (Da R2026a)
featureSelectionClassificationKruskalWallisComponentPipeline component for performing feature selection using Kruskal-Wallis test (Da R2026a)
featureSelectionClassificationMRMRComponentPipeline component for performing MRMR feature selection in classification workflow (Da R2026a)
featureSelectionClassificationNCAComponentPipeline component for performing feature selection using neighborhood component analysis (NCA) for classification (Da R2026a)
featureSelectionClassificationReliefFComponentPipeline component for performing feature selection using ReliefF algorithm (Da R2026a)
featureSelectionRegressionFTestComponentPipeline component for performing feature selection using F-tests (Da R2026a)
featureSelectionRegressionMRMRComponentPipeline component for performing MRMR feature selection in regression workflow (Da R2026a)
featureSelectionRegressionNCAComponentPipeline component for performing feature selection using neighborhood component analysis (NCA) for regression (Da R2026a)
featureSelectionRegressionReliefFComponentPipeline component for performing feature selection using RReliefF algorithm (Da R2026a)
variableSelectorComponentPipeline component for manual variable selection (Da R2026a)

Componenti di classificazione

classificationDiscriminantComponentPipeline component for discriminant analysis classification (Da R2026a)
classificationECOCComponentPipeline component for multiclass classification using error-correcting output codes (ECOC) model (Da R2026a)
classificationEnsembleComponentPipeline component for ensemble classification (Da R2026a)
classificationGAMComponentPipeline component for binary classification using generalized additive model (GAM) (Da R2026a)
classificationKernelComponentPipeline component for classification using Gaussian kernel with random feature expansion (Da R2026a)
classificationKNNComponentPipeline component for classification using k-nearest neighbor model (Da R2026a)
classificationLinearComponentPipeline component for binary classification of high-dimensional data using linear model (Da R2026a)
classificationNaiveBayesComponentPipeline component for multiclass classification using naive Bayes model (Da R2026a)
classificationNeuralNetworkComponentPipeline component for classification using neural network model (Da R2026a)
classificationSVMComponentPipeline component for one-class and binary classification using SVM classifier (Da R2026a)
classificationTreeComponentPipeline component for multiclass classification using binary decision trees (Da R2026a)

Componenti di regressione

regressionEnsembleComponentPipeline component for regression using ensemble of learners (Da R2026a)
regressionGAMComponentPipeline component for generalized additive model (GAM) for regression (Da R2026a)
regressionGPComponentPipeline component for Gaussian process regression (GPR) (Da R2026a)
regressionLinearComponentPipeline component for regression of high-dimensional data using a linear model (Da R2026a)
regressionKernelComponentPipeline component for regression using Gaussian kernel model (Da R2026a)
regressionNeuralNetworkComponentPipeline component for regression using neural network model (Da R2026a)
regressionSVMComponentPipeline component for regression using a support vector machine (SVM) model (Da R2026a)
regressionTreeComponentPipeline component for regression using binary decision trees (Da R2026a)
functionComponentPipeline component for custom function (Da R2026a)

Funzioni

espandi tutto

Collegamento automatico

seriesConnect components in series to create pipeline (Da R2026a)
parallelConnect components or pipelines in parallel to create pipeline (Da R2026a)
insertInsert component or pipeline into existing pipeline (Da R2026a)
replaceReplace existing pipeline component with new component (Da R2026a)

Collegamento manuale

addAdd new component or pipeline to existing pipeline (Da R2026a)
removeRemove existing components or pipelines from pipeline (Da R2026a)
connectCreate connections between pipeline components (Da R2026a)
disconnectRemove connections between ports in pipeline (Da R2026a)

Gerarchia

expandExpand subpipelines in pipeline (Da R2026a)
learnInitialize and evaluate pipeline or component (Da R2026a)
runExecute pipeline or component for inference after learning (Da R2026a)
pruneRemove unnecessary components and dependencies from pipeline (Da R2026a)
resetReset pipeline or component (Da R2026a)
crossvalidateCross-validate pipeline (Da R2026a)
viewView diagram of pipeline inputs, outputs, components, and connections (Da R2026a)
describeDisplay summary of pipeline components (Da R2026a)
packageCreate deployable archive or standalone application from pipeline (Da R2026a)

Argomenti

Esempi in primo piano

Create Simple Classification Pipeline

Create Simple Classification Pipeline

Create, learn, and run a machine learning pipeline for SVM classification.

Tune Pipeline Hyperparameters Using Cross-Validation

Tune Pipeline Hyperparameters Using Cross-Validation

Use cross-validation to select a pipeline parameter value.