Scikit-learn Snippets
Important notice: This is a pre-release extension where some of the advertised features are not yet complete, and those features that do exist are subject to change.
This extension is for data scientists writing Python code to create, fit, and evaluate machine learning models using thescikit-learn package. It provides code snippets that work with Python (.py) files or (.ipynb) notebooks to ease the friction of working with scikit-learn. Snippets boost productivity by reducing the amount of Python syntax and function arguments that you need to remember, eliminating keystrokes, and helping to free the data scientist to focus on building useful models from their data.
1. Usage
1.1 Overview
All snippets provided by this extension have triggers prefixed sk, which activate the IntelliSense pop up or allow filtering in the Command Palette. Trigger prefixes are further organised as follows:
| Prefix |
Description |
sk-setup |
Starting point for importing commonly used modules and setting defaults. |
sk-read |
Read input training data or existing models from file. |
sk-prep |
Preprocess input training data for model fitting. |
sk-regress |
Create and fit regression models. |
sk-classify |
Create and fit classification models. |
sk-cluster |
Create and fit clustering models. |
sk-density |
Create and fit density estimation models. |
sk-embed |
Create and fit dimensionality reduction (embedding) models. |
sk-anomaly |
Create and fit anomaly detection models. |
sk-validation |
Model validation. |
sk-inspect |
Model inspection and explainability. |
sk-io |
Save and restore models on disk. |
sk-args |
Select and adjust model parameters from lists of valid options. |
|
|
For more information about the organization of the snippets provided by this extension see Section 4.2 for a visual overview of the complete hierarchy.
1.2 Snippets for machine learning workflows
A typical workflow is to:
import commonly used modules with sk-setup,
read training data from file with sk-read,
preprocess training data with sk-prep,
create and train models across a range of machine learning tasks such as:
regression (sk-regress),
classification (sk-classify),
clustering (sk-cluster),
density estimation (sk-density),
dimensionality reduction (sk-embed),
anomaly detection (sk-anomaly)
evaluate fitted models by cross-validation (sk-validation) and inspection (sk-inspect),
save and restore models on disk (sk-io),
optionally adjust the parameters that control model fitting with sk-args.
See the Features section below for a full list of the available snippets and their prefix triggers.
1.3 Inserting snippets
Inserting machine learning code snippets into your Python code is easy. Use either of these methods:
Command Palette
Click inside a Python notebook cell or editor and choose Insert Snippet from the Command Palette.
A list of snippets appears. You can type to filter the list; start typing sk to filter the list for snippets provided by this extension. You can further filter the list by continuing to type the desired prefix.
Choose a snippet from the list and it is inserted into your Python code.
Placeholders indicate the minimum arguments required to train a model. Use the tab key to step through the placeholders.
IntelliSense
Start typing sk in a Python notebook cell or editor.
The IntelliSense pop up will appear. You can further filter the pop up list by continuing to type the desired prefix.
Choose a snippet from the pop up and it is inserted into your Python code.
Placeholders indicate the minimum arguments required to train a model. Use the tab key to step through the placeholders.
You can also trigger IntelliSense by typing Ctrl+Space.
2. Features
2.1 Setup
The following snippets are triggered by sk-setup and provide the starting point for creating models. Usually inserted near the beginning of a code file or notebook, sk-setup is the key snippet for importing modules that are commonly used in the machine learning workflow, and setting defaults that apply to data visualizations.
| Snippet |
Placeholders |
Description |
sk-setup |
pio.renderers.default
pio.templates.default |
Provides the initial starting point for creating models. Imports commonly used modules (pandas, numpy, json, pickle, plotly.express, and plotly.io), and sets the default figure renderer and template. |
|
|
|
2.2 Read training data
The following snippets are triggered by sk-read and provide features for creating pandas data frames from Comma Separated Value (.csv), Microsoft Excel (.xlsx), Feather (.feather), and Parquet (.parquet) format files. Tabular data stored in pandas data frames is a common source of training data required for fitting scikit-learn models.
| Snippet |
Placeholders |
Description |
sk-read-csv |
df,
file |
Read tabular training data from CSV (.csv) file (file) into pandas data frame (df) and report info. |
sk-read-excel |
df,
file |
Read tabular training data from Excel (.xlsx) file (file) into pandas data frame (df) and report info. |
sk-read-feather |
df,
file |
Read tabular training data from Feather (.feather) file (file) into pandas data frame (df) and report info. |
sk-read-parquet |
df,
file |
Read tabular training data from Parquet (.parquet) file (file) into pandas data frame (df) and report info. |
|
|
|
2.3 Preprocessing for supervised learning
The following snippets are triggered by sk-prep-target and provide features for preparing tabular training data for supervised learning. The data preparation process involves extracting one or more named features (X) and a named target variable (y) from an input data frame (df), and creating an input training dataset. The data arrays are used to train models for a range of supervised machine learning tasks.
Optionally, the process can also include extracting one or more secondary variables (Z) that are used to better understand the results of model fitting. It is important to note that these secondary variables do not play any role in the model fitting itself, but they can be useful for interpreting the results.
| Snippet |
Placeholders |
Description |
sk-prep-target-features |
X1, X2, X3,
Y
df,
X, y |
Prepare training data (X, y) for supervised learning. Training data is identified by a sequence of feature names (X1, X2, X3, ...), and a target name (Y), sourced from a data frame (df). |
sk-prep-target-features-secondary |
X1, X2, X3,
Y
Z1, Z2, Z3,
df,
X, y, Z |
Prepare training data (X, y) for supervised learning, and prepare secondary data (Z) for model evaluation. Training data is identified by a sequence of feature names (X1, X2, X3, ...), and a target name (Y), sourced from a data frame (df). Secondary data is also identified by a sequence of feature names (Z1, Z2, Z3, ...) sourced from the same data frame (df).
Note: Secondary data is used only for interpreting model output and plays no role in model training. |
sk-prep-train_test_split |
X, y,
⚙ train_size,
⚙ random_state |
Randomly split input data (X, y) into training and test sets using the train_test_split function with the supplied parameters (⚙). Holding out a test set from the training process helps to evaluate the performance of supervised learning models. |
|
|
|
2.4 Preprocessing for unsupervised learning
The following snippets are triggered by sk-prep-features and provide features for preparing tabular training data for unsupervised learning. The data preparation process involves extracting one or more named features (X) from an input data frame (df), and creating an input training dataset. The data array is used to train models for a range of unsupervised machine learning tasks. In contrast to supervised learning, there is no target variable (y) in unsupervised learning.
Similar to the supervised learning case, the process can optionally include extracting one or more secondary variables (Z) that are used to better understand the results of model fitting. It is important to note that these secondary variables do not play any role in the model fitting itself, but they can be useful for interpreting the results.
| Snippet |
Placeholders |
Description |
sk-prep-features |
X1, X2, X3
df,
X |
Prepare training data features (X) for unsupervised learning. Training data is identified by a sequence of feature names (X1, X2, X3, ...) sourced from a data frame (df). |
sk-prep-features-secondary |
X1, X2, X3,
Z1, Z2, Z3,
df,
X, Z |
Prepare training data features (X) for unsupervised learning, and prepare secondary data (Z) for model evaluation. Training data is identified by a sequence of feature names (X1, X2, X3, ...) sourced from a data frame (df). Secondary data is also identified by a sequence of feature names (Z1, Z2, Z3, ...) sourced from the same data frame (df).
Note: Secondary data is used only for interpreting model output and plays no role in model training. |
|
|
|
2.5 Regression
2.5.1 Linear regression
The following snippets are triggered by sk-regress-linear and provide features for various types of linear regression ranging from simple ordinary least squares (LinearRegression), regression with a transformed target (TransformedTargetRegressor), regression with transformed features (FunctionTransformer, PolynomialFeatures, SplineTransformer), and regularized models such as ridge regression (Ridge), lasso regression (Lasso), and elastic net regression (ElasticNet).
| Snippet |
Placeholders |
Description |
sk-regress-linear |
estimator_linear,
⚙ fit_intercept,
⚙ positive,
X, y |
Linear regression: Create and fit a LinearRegression regression model (estimator_linear) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-transform-target |
estimator_transform_target,
⚙ func,
⚙ func_inverse,
⚙ fit_intercept,
⚙ positive,
⚙ check_inverse,
X, y |
Linear regression with transformed target: Create and fit a TransformedTargetRegressor regression model (estimator_transform_target) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-transform |
estimator_transform,
⚙ func,
⚙ fit_intercept,
⚙ positive,
X, y |
Linear regression with transformed features: Create and fit a LinearRegression with FunctionTransformer regression model (estimator_transform) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-polynomial |
estimator_polynomial,
⚙ degree,
⚙ fit_intercept,
⚙ positive,
X, y |
Polynomial regression: Create and fit a LinearRegression with PolynomialFeatures regression model (estimator_polynomial) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-spline |
estimator_spline,
⚙ n_knots,
⚙ degree,
⚙ knots,
⚙ extrapolation,
⚙ fit_intercept,
⚙ positive,
X, y |
Spline regression: Create and fit a LinearRegression with SplineTransformer regression model (estimator_spline) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-pcr |
estimator_pcr,
⚙ n_components,
⚙ whiten,
⚙ fit_intercept,
⚙ positive,
X, y |
Principal component regression (PCR): Create and fit a LinearRegression with PCA regression model (estimator_pcr) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-pls |
estimator_pls,
⚙ n_components,
⚙ scale,
⚙ max_iter,
X, y |
Partial least squares regression (PLS): Create and fit a PLSRegression regression model (estimator_pls) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-ridge |
estimator_ridge,
⚙ alpha,
⚙ fit_intercept,
⚙ positive,
X, y |
Ridge regression: Create and fit a Ridge regression model (estimator_ridge) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-ridgecv |
estimator_ridgecv,
⚙ alphas,
⚙ cv,
⚙ fit_intercept,
X, y |
Ridge regression with cross-validation: Create and fit a RidgeCV regression model (estimator_ridgecv) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-lasso |
estimator_lasso,
⚙ alpha,
⚙ fit_intercept
⚙ positive,
⚙ selection,
⚙ random_state,
X, y |
Lasso regression: Create and fit a Lasso regression model (estimator_lasso) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-lassocv |
estimator_lassocv,
⚙ alphas,
⚙ cv,
⚙ fit_intercept,
⚙ positive,
⚙ selection,
⚙ random_state,
X, y |
Lasso regression with cross-validation: Create and fit a LassoCV regression model (estimator_lassocv) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-elasticnet |
estimator_elasticnet,
⚙ alpha,
⚙ l1_ratio,
⚙ fit_intercept,
⚙ positive,
⚙ selection,
⚙ random_state,
X, y |
ElasticNet regression: Create and fit an ElasticNet regression model (estimator_elasticnet) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-linear-elasticnetcv |
estimator_elasticnetcv,
⚙ l1_ratio,
⚙ alphas,
⚙ fit_intercept,
⚙ positive,
⚙ selection,
⚙ random_state,
X, y |
ElasticNet regression with cross-validation: Create and fit an ElasticNetCV regression model (estimator_elasticnetcv) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-dummy |
estimator_dummy,
⚙ strategy,
⚙ constant,
⚙ quantile,
X, y |
Dummy regression: Create and fit a DummyRegressor regression model (estimator_dummy) with the supplied parameters (⚙) and training data (X, y). |
|
|
|
2.5.2 Ensemble regression
The following snippets are triggered by sk-regress-ensemble and provide features for various types of ensemble regression. Ensemble regression models combine multiple base regression models to create a model that has superior performance compared to a single base model. Rather than relying on a single model's prediction, ensemble methods aggregate predictions from several models to produce a final result that is typically more accurate.
| Snippet |
Placeholders |
Description |
sk-regress-ensemble-random-forest |
estimator_random_forest,
⚙ n_estimators,
⚙ criterion,
⚙ min_samples_leaf,
⚙ random_state,
X, y |
Random forest regression: Create and fit a RandomForestRegressor regression model (estimator_random_forest) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-ensemble-extra-trees |
estimator_extra_trees,
⚙ n_estimators,
⚙ criterion,
⚙ min_samples_leaf,
⚙ random_state,
X, y |
Extremely randomized trees (extra-trees) regression: Create and fit an ExtraTreesRegressor regression model (estimator_extra_trees) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-ensemble-gradient-boosting |
estimator_gradient_boosting,
⚙ n_estimators,
⚙ loss,
⚙ learning_rate,
⚙ min_samples_leaf,
⚙ random_state,
X, y |
Gradient boosting regression: Create and fit a GradientBoostingRegressor regression model (estimator_gradient_boosting) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-ensemble-hist-gradient-boosting |
estimator_hist_gradient_boosting,
⚙ loss,
⚙ learning_rate,
⚙ min_samples_leaf,
⚙ random_state,
X, y |
Histogram-based gradient boosting regression: Create and fit a HistGradientBoostingRegressor regression model (estimator_hist_gradient_boosting) with the supplied parameters (⚙) and training data (X, y). |
sk-regress-ensemble-stacking |
estimator_stacking,
⚙ estimators,
⚙ final_estimator,
⚙ cv,
⚙ passthrough,
X, y |
Stack of estimators with a final regressor: Create and fit a StackingRegressor regression model (estimator_stacking) with the supplied parameters (⚙) and training data (X, y).
Note: By default, all regression estimators in the current scope are collected for stacking. |
|
|
|
2.6 Classification
TO BE COMPLETED
The following snippets are triggered by sk-classify and provide features for various types of classification models. Classification is a supervised learning task where the goal is to predict discrete class labels or categories for input data. These models learn from labeled training data and can be used for binary classification (two classes) or multi-class classification problems.
| Snippet |
Placeholders |
Description |
sk-classify-lda |
n_components,
whiten |
Linear Discriminant Analysis (LDA): Create and fit a LinearDiscriminantAnalysis classification model (classifier_lda) with the supplied parameters (⚙) and training data (X, y). |
sk-classify-qda |
n_components,
kernel,
gamma,
degree,
coef0 |
Quadratic Discriminant Analysis (QDA): Create and fit a QuadraticDiscriminantAnalysis classification model (classifier_qda) with the supplied parameters (⚙) and training data (X, y). |
|
|
|
2.7 Clustering
The following snippets are triggered by sk-cluster and provide features for various types of clustering models. Clustering is an unsupervised learning task that groups similar data points together based on their features. These models identify natural groupings within a dataset without requiring labeled examples, making them useful for discovering patterns, segmenting data, and identifying structure in unlabeled datasets.
| Snippet |
Placeholders |
Description |
sk-cluster-kmeans |
estimator_kmeans,
⚙ n_clusters,
⚙ init,
⚙ n_init,
⚙ max_iter,
⚙ tol,
⚙ random_state,
⚙ algorithm,
X |
K-means clustering: Create and fit a KMeans clustering model (estimator_kmeans) with the supplied parameters (⚙) and training data (X). Includes StandardScaler in pipeline for proper feature scaling. |
sk-cluster-kmeans-minibatch |
estimator_kmeans_minibatch,
⚙ n_clusters,
⚙ init,
⚙ batch_size,
⚙ max_iter,
⚙ max_no_improvement,
⚙ tol,
⚙ random_state,
⚙ reassignment_ratio,
X |
MiniBatch K-means clustering: Create and fit a MiniBatchKMeans clustering model (estimator_kmeans_minibatch) with the supplied parameters (⚙) and training data (X). Optimized for large datasets with StandardScaler in pipeline. |
sk-cluster-meanshift |
estimator_meanshift,
⚙ bandwidth,
⚙ seeds,
⚙ bin_seeding,
⚙ min_bin_freq,
⚙ cluster_all,
⚙ max_iter,
X |
Mean shift clustering: Create and fit a MeanShift clustering model (estimator_meanshift) with the supplied parameters (⚙) and training data (X). Includes StandardScaler in pipeline for proper feature scaling. |
sk-cluster-dbscan |
estimator_dbscan,
⚙ eps,
⚙ min_samples,
⚙ metric,
⚙ algorithm,
⚙ leaf_size,
X |
DBSCAN clustering: Create and fit a DBSCAN clustering model (estimator_dbscan) with the supplied parameters (⚙) and training data (X). Includes StandardScaler in pipeline which is critical for distance-based algorithms. |
sk-cluster-hdbscan |
estimator_hdbscan,
⚙ min_cluster_size,
⚙ min_samples,
⚙ cluster_selection_epsilon,
⚙ cluster_selection_method,
⚙ alpha,
⚙ metric,
⚙ alpha,
⚙ algorithm,
⚙ leaf_size,
X |
HDBSCAN clustering: Create and fit an HDBSCAN clustering model (estimator_hdbscan) with the supplied parameters (⚙) and training data (X). |
sk-cluster-predict |
estimator,
X |
Cluster prediction: Apply a clustering model (estimator) to an input dataset (X) to predict cluster labels using the model's predict() function. Cluster labels are output to a new dataset (X_estimator_cluster). Only available for models that support prediction on new data. |
|
|
|
2.8 Density estimation
The following snippets are triggered by sk-density and provide features for various types of density estimation models. Density estimation is an unsupervised learning task that creates a model of the probability distribution from which the observed data is drawn. These models can be used to generate new samples, detect outliers, and estimate the likelihood of data points.
| Snippet |
Placeholders |
Description |
sk-density-kernel |
estimator_kernel_density,
⚙ bandwidth,
⚙ kernel,
⚙ metric,
X |
Kernel Density Estimation: Create and fit a KernelDensity density estimation model (estimator_kernel_density) with the supplied parameters (⚙) and training data (X). |
sk-density-gaussian-mixture |
estimator_gaussian_mixture,
⚙ n_components,
⚙ covariance_type,
⚙ init_params,
⚙ random_state,
⚙ max_iter,
X |
Gaussian Mixture Model: Create and fit a GaussianMixture density estimation model (estimator_gaussian_mixture) with the supplied parameters (⚙) and training data (X). |
sk-density-sample-kernel |
estimator,
⚙ n_samples,
⚙ random_state, |
Sample from Kernel Density model: Generate random samples from a fitted kernel density model (estimator) using the sample() function with the supplied parameters (⚙). Samples are output to a new dataset (estimator_samples). |
sk-density-sample-gaussian-mixture |
estimator,
⚙ n_samples, |
Sample from Gaussian Mixture model: Generate random samples from a fitted Gaussian Mixture density model (estimator) using the sample() function with the supplied parameters (⚙). Samples are output to new datasets (estimator_samples, and estimator_components). |
sk-density-score-samples |
estimator,
X |
Density score of each sample: Apply a density estimation model (estimator) to an input dataset (X) to evaluate the log-likelihood of each sample using the model's score_samples() function. The log-likelihood is output to a new dataset (X_estimator_density), and is normalized to be a probability density, so the value will be low for high-dimensional data. |
sk-density-score |
estimator,
X |
Density score: Apply a density estimation model (estimator) to an input dataset (X) to evaluate the total log-likelihood of the data in X using the model's score() function. This is normalized to be a probability density, so the value will be low for high-dimensional data. |
|
|
|
2.9 Dimensionality reduction
The following snippets are triggered by sk-embed and provide features for various types of dimensionality reduction or embedding. Dimensionality reduction algorithms transform data from a high-dimensional space into a lower-dimensional representation while preserving the most important structure or information. These techniques are useful for visualization, computational efficiency, and removing redundant features.
| Snippet |
Placeholders |
Description |
sk-embed-pca |
estimator_pca,
⚙ n_components,
⚙ whiten
X |
Principal Component Analysis: Create and fit a PCA dimensionality reduction model (estimator_pca) with the supplied parameters (⚙) and training data (X). |
sk-embed-kpca |
estimator_kpca,
⚙ n_components,
⚙ kernel,
⚙ gamma,
⚙ degree,
⚙ coef0
X |
Kernel PCA: Create and fit a KernelPCA dimensionality reduction model (estimator_kpca) with the supplied parameters (⚙) and training data (X). |
sk-embed-lle |
estimator_lle,
⚙ n_components,
⚙ n_neighbors,
⚙ method
X |
Locally Linear Embedding: Create and fit a LocallyLinearEmbedding dimensionality reduction model (estimator_lle) with the supplied parameters (⚙) and training data (X). |
sk-embed-isomap |
estimator_isomap,
⚙ n_components,
⚙ n_neighbors,
⚙ radius,
⚙ p
X |
Isometric Mapping: Create and fit an Isomap dimensionality reduction model (estimator_isomap) with the supplied parameters (⚙) and training data (X). |
sk-embed-mds |
estimator_mds,
⚙ n_components,
⚙ metric,
⚙ n_init,
⚙ random_state,
⚙ normalized_stress
X |
Multidimensional Scaling: Create and fit an MDS dimensionality reduction model (estimator_mds) with the supplied parameters (⚙) and training data (X). |
sk-embed-spectral |
estimator_spectral,
⚙ n_components,
⚙ affinity,
⚙ gamma,
⚙ random_state,
⚙ n_neighbors
X |
Spectral Embedding: Create and fit a SpectralEmbedding dimensionality reduction model (estimator_spectral) with the supplied parameters (⚙) and training data (X). |
sk-embed-tsne |
estimator_tsne,
⚙ n_components,
⚙ perplexity,
⚙ random_state,
⚙ n_iter
X |
t-Distributed Stochastic Neighbour Embedding: Create and fit a TSNE dimensionality reduction model (estimator_tsne) with the supplied parameters (⚙) and training data (X). Embedding is output to X_estimator_tsne. |
sk-embed-transform |
estimator,
X |
Embedding transform: Apply a dimensionality reduction model (estimator) to an input dataset (X) using the model's transform() function. Embedding is output to a new dataset (X_estimator). |
|
|
|
2.10 Anomaly detection
The following snippets are triggered by sk-anomaly and provide features for various types of anomaly detection models. Anomaly detection is the task of identifying rare items, events, or observations that differ significantly from the majority of the data. These models learn patterns in the data and can detect unusual instances that do not conform to expected behavior.
| Snippet |
Placeholders |
Description |
sk-anomaly-one-class-svm |
estimator_one_class_svm
⚙ kernel,
⚙ gamma,
⚙ nu,
⚙ shrinking,
X |
One-Class Support Vector Machine: Create and fit a OneClassSVM anomaly detection model (estimator_one_class_svm) with the supplied parameters (⚙) and training data (X). |
sk-anomaly-one-class-svm-sgd |
estimator_one_class_svm_sgd
⚙ kernel,
⚙ gamma,
⚙ n_components,
⚙ random_state,
⚙ nu,
⚙ fit_intercept,
⚙ max_iter,
⚙ tol,
⚙ shuffle,
⚙ learning_rate,
⚙ eta0,
X |
One-Class SVM with Stochastic Gradient Descent: Create and fit a SGDOneClassSVM with Nystroem kernel anomaly detection model (estimator_one_class_svm_sgd) with the supplied parameters (⚙) and training data (X). |
sk-anomaly-local-outlier-factor |
estimator_lof
⚙ n_neighbors,
⚙ algorithm,
⚙ leaf_size,
⚙ metric,
⚙ contamination,
⚙ novelty,
X |
Local Outlier Factor: Create and fit a LocalOutlierFactor anomaly detection model (estimator_lof) with the supplied parameters (⚙) and training data (X). |
sk-anomaly-isolation-forest |
estimator_isolation_forest
⚙ n_estimators,
⚙ max_samples,
⚙ contamination,
⚙ max_features,
⚙ bootstrap,
⚙ n_jobs,
⚙ random_state,
X |
Isolation Forest: Create and fit an IsolationForest anomaly detection model (estimator_isolation_forest) with the supplied parameters (⚙) and training data (X). |
sk-anomaly-elliptic-envelope |
estimator_elliptic_envelope
⚙ store_precision,
⚙ assume_centered,
⚙ support_fraction,
⚙ contamination,
⚙ random_state,
X |
Elliptic Envelope (Robust Covariance): Create and fit an EllipticEnvelope anomaly detection model (estimator_elliptic_envelope) with the supplied parameters (⚙) and training data (X). |
sk-anomaly-dbscan |
estimator_dbscan
⚙ eps,
⚙ min_samples,
⚙ metric,
⚙ algorithm,
⚙ leaf_size,
X |
DBSCAN: Create and fit a DBSCAN anomaly detection model (estimator_dbscan) with the supplied parameters (⚙) and training data (X). |
sk-anomaly-predict |
estimator,
X |
Anomaly prediction: Apply an anomaly detection model (estimator) to an input dataset (X) to perform outlier classification using the model's predict() function. Outlier class is output to a new dataset (X_estimator_class) where -1 indicates outliers, and +1 indicates inliers. |
sk-anomaly-score |
estimator,
X |
Anomaly score: Apply an anomaly detection model (estimator) to an input dataset (X) to evaluate the outlier score of each sample using the model's descision_function() function. Outlier score is output to a new dataset (X_estimator_score) where negative scores indicate outliers, and positive scores indicate inliers. |
|
|
|
2.11 Model inspection
The following snippets are triggered by sk-inspect and provide features for inspecting and understanding fitted models. Model inspection tools help data scientists gain insights into how their models make predictions, which features are most important, and how changes in input features affect model outputs. These techniques are essential for explaining model behavior, debugging models, and ensuring that models are behaving as expected.
| Snippet |
Placeholders |
Description |
sk-inspect-partial_dependence |
estimator,
X,
⚙ features,
⚙ percentiles,
⚙ grid_resolution,
⚙ kind |
Partial dependence: Compute the partial dependence of a model (estimator) against an input feature dataset (X) using the partial_dependence function with the supplied parameters (⚙). Output partial dependence curves are returned as a dictionary (estimator_partial). |
sk-inspect-permutation_importance |
estimator,
X, y,
⚙ scoring,
⚙ n_repeats,
⚙ random_state,
⚙ sample_weight,
⚙ max_samples |
Permutation importance: Compute the permutation importance of a model (estimator) against an input dataset (X,y) using the permutation_importance function with the supplied parameters (⚙). Output feature importance metrics are returned as a dictionary (estimator_permutation). |
|
|
|
2.12 Model storage
The following snippets are triggered by sk-io and provide features for reading and writing scikit-learn models. The pickle module is used to read and write scikit-learn models because it provides the required binary file i/o capabilities.
Note: There are security risks involved in reading pickle binary files, and users should only read models from trusted sources.
| Snippet |
Placeholders |
Description |
sk-io-read-pickle |
file,
estimator |
Read (or deserialize) an existing model (estimator) from a binary (.pickle) format file (file). Report the model type and any fitted parameters. |
sk-io-write-pickle |
file,
estimator |
Write (or serialize) a model (estimator) to a binary (.pickle) format file (file). |
|
|
|
2.13 Argument Snippets
Many scikit-learn function arguments take their values from extensive option lists. The following snippets are triggered with sk-args and provide features allowing you to easily select valid argument values from lists of available options.
| Snippet |
Placeholders |
Description |
sk-args-random_state |
random_state |
Set the random_state argument for reproducibility in randomized algorithms. This argument is the integer seed number used to initialize random number generation; a randomly chosen value is provided by default. |
sk-args-alphas |
logspace or linspace,
start, stop, num |
Set the alphas argument (a logarithmic or linear sequence of regularization parameters) for cross-validation in RidgeCV, LassoCV, and ElasticNetCV regression. |
sk-args-func |
func |
Set the func argument to a FunctionTransformer from a list of common transformations. For use in regression models with transformed features (X), see sk-regress-linear-transform. |
sk-args-func-inverse |
func, inverse_func |
Set the func and inverse_func arguments to a FunctionTransformer from a list of common forward and inverse transformation function pairs. For use in regression models with transformed target (y), see sk-regress-linear-transform-target. |
sk-args-spline-extrapolation |
extrapolation |
Set the SplineTransformer extrapolation behavior beyond the minimum and maximum of the training data, see sk-regress-linear-spline. |
sk-args-kernel |
kernel |
Set the kernel type for kernel density estimation from an option list. You can provide this argument when creating a KernelDensity model, see sk-density-kernel. |
|
|
|
3. Release Notes
3.1 Scikit-learn version
The snippets provided by this extension were developed using scikit-learn version 1.7 but will also produce working Python code for earlier and later versions.
3.2 Editor Support for Snippets
Snippets for producing Python code, including those provided by this extension, are supported in the Python file (.py) editor and in the notebook (.ipynb) editor.
3.3 Snippets and IntelliSense
When triggered, the default behaviour of IntelliSense is to show snippets along with other context dependent suggestions. This may result in a long list of suggestions in the IntelliSense pop-up, particularly if the snippet trigger provided by this extension (sk) also matches other symbols in your editor.
It's easy to modify this behaviour using your Visual Studio Code settings. To access the relevant settings go to Preferences > Settings and type snippets in the Search settings field as shown below:
You can control whether snippets are shown with other suggestions and how they are sorted using the Editor: Snippet Suggestions dropdown. Choose one of the options to control how snippet suggestions are shown in the IntelliSense popup:
| Option |
IntelliSense |
top |
Show snippet suggestions on top of other suggestions. |
bottom |
Show snippet suggestions below other suggestions. |
inline |
Show snippet suggestions with other suggestions (default). |
none |
Do not show snippet suggestions. |
|
|
You can also use the Editor > Suggest: Show Snippets checkbox to enable or disable snippets in IntelliSense suggestions. When snippets are disabled in IntelliSense they are still accessible through the Command Palette Insert Snippet command.
4. Reference
4.1 Scikit-learn documentation
For more information about scikit-learn including extensive documentation and examples visit https://scikit-learn.org/stable/index.html
4.2 Scikit-learn snippet tree
Snippet prefix triggers are organized in a hierarchical tree structure rooted at sk as shown in the figure below. The snippet hierarchy is designed to ease the user's cognitive load when developing models with this large and complex machine learning package. The branches at the top of the tree outline the main steps in a machine learning workflow, branches at lower levels outline a taxonomy of algorithms for specific tasks, whereas leaf nodes represent particular algorithms. The process of inserting a snippet amounts to navigating the tree and selecting the desired leaf node by either of the methods described in Section 1.3.
sk
├── sk-setup
├── sk-read
│ ├── sk-read-csv
│ ├── sk-read-excel
│ ├── sk-read-feather
│ └── sk-read-parquet
├── sk-prep
│ ├── sk-prep-target-features
│ ├── sk-prep-target-features-secondary
│ ├── sk-prep-train_test_split
│ ├── sk-prep-features
│ └── sk-prep-features-secondary
├── sk-regress
│ ├── sk-regress-linear
│ │ ├── sk-regress-linear
│ │ ├── sk-regress-linear-transform-target
│ │ ├── sk-regress-linear-transform
│ │ ├── sk-regress-linear-polynomial
│ │ ├── sk-regress-linear-spline
│ │ ├── sk-regress-linear-pcr
│ │ ├── sk-regress-linear-pls
│ │ ├── sk-regress-linear-ridge
│ │ ├── sk-regress-linear-ridgecv
│ │ ├── sk-regress-linear-lasso
│ │ ├── sk-regress-linear-lassocv
│ │ ├── sk-regress-linear-elasticnet
│ │ └── sk-regress-linear-elasticnetcv
│ ├── sk-regress-ensemble
│ │ ├── sk-regress-ensemble-random-forest
│ │ ├── sk-regress-ensemble-extra-trees
│ │ ├── sk-regress-ensemble-gradient-boosting
│ │ ├── sk-regress-ensemble-hist-gradient-boosting
│ │ └── sk-regress-ensemble-stacking
│ └── sk-regress-dummy
├── sk-cluster
│ ├── sk-cluster-kmeans
│ ├── sk-cluster-kmeans-minibatch
│ ├── sk-cluster-meanshift
│ ├── sk-cluster-dbscan
│ ├── sk-cluster-hdbscan
│ └── sk-cluster-predict
├── sk-density
│ ├── sk-density-kernel
│ ├── sk-density-gaussian-mixture
│ ├── sk-density-sample-kernel
│ ├── sk-density-sample-gaussian-mixture
│ ├── sk-density-score-samples
│ └── sk-density-score
├── sk-embed
│ ├── sk-embed-pca
│ ├── sk-embed-kpca
│ ├── sk-embed-lle
│ ├── sk-embed-isomap
│ ├── sk-embed-mds
│ ├── sk-embed-spectral
│ ├── sk-embed-tsne
│ └── sk-embed-transform
├── sk-anomaly
│ ├── sk-anomaly-one-class-svm
│ ├── sk-anomaly-one-class-svm-sgd
│ ├── sk-anomaly-local-outlier-factor
│ ├── sk-anomaly-isolation-forest
│ ├── sk-anomaly-elliptic-envelope
│ ├── sk-anomaly-dbscan
│ ├── sk-anomaly-predict
│ └── sk-anomaly-score
├── sk-inspect
│ ├── sk-inspect-partial_dependence
│ └── sk-inspect-permutation_importance
├── sk-io
│ ├── sk-io-read-pickle
│ └── sk-io-write-pickle
└── sk-args
├── sk-args-random_state
├── sk-args-alphas
├── sk-args-func
├── sk-args-func-inverse
├── sk-args-spline-extrapolation
└── sk-args-kernel
Copyright © 2024-2025 Analytic Signal Limited, all rights reserved
| |
