Driverless AI Transformations

Transformations in Driverless AI are applied to columns in the data. The transformers create the engineered features in experiments.

Driverless AI provides a number of transformers. The downloaded experiment logs include the transformations that were applied to your experiment.

Notes:

  • You can include or exclude specific transformers in your Driverless AI environment using the included_transformers or excluded_transformers config options.

  • You can control which transformers to use in individual experiments with the included_transformers Expert Setting in Recipe panel.

  • You can set transformers to be used as pre-processing transformers with the included_pretransformers Expert Setting in Recipe panel. Additional layers can be added with the num_pipeline_layers Expert Setting in Recipe panel.

  • An alternative to transformers that gives more flexibility (but has no fitted state) are data recipes, controlled by the included_datas Expert Setting in Recipe panel.

Available Transformers

The following transformers are available for regression and classification (multiclass and binary) experiments.

Transformed Feature Naming Convention

Transformed feature names are encoded as follows:

<Transformation_indexORgene_details_id>_<Transformation_name>:<original_feature_name>:<…>:<original_feature_name>.<extra>

For example in 32_NumToCatTE:BILL_AMT1:EDUCATION:MARRIAGE:SEX.0 :

  • 32_ is the transformation index for specific transformation parameters.

  • NumToCatTE is the transformer name.

  • BILL_AMT1:EDUCATION:MARRIAGE:SEX represents original features used.

  • 0 is the extra and represents the likelihood encoding for target[0] after grouping by features (shown here as BILL_AMT1, EDUCATION, MARRIAGE and SEX) and making out-of-fold estimates. For multiclass experiments, this value is > 0. For binary experiments, this value is always 0.

Numeric Transformers (Integer, Real, Binary)

  • ClusterDist Transformer

    The Cluster Distance Transformer clusters selected numeric columns and uses the distance to a specific cluster as a new feature.

  • ClusterDist cuML Transformer

    The Cluster Distance cuML Transformer runs on GPUs to train cuML accelerated k-means clustering to create clusters on selected numeric columns and uses the distance to a specific cluster as a new feature.

  • ClusterTE Transformer

    The Cluster Target Encoding Transformer clusters selected numeric columns and calculates the mean of the response column for each cluster. The mean of the response is used as a new feature. Cross Validation is used to calculate mean response to prevent overfitting.

  • DBSCAN cuML Transformer

    DBSCAN cuML Transformer runs on GPUs to train cuML accelerated DBSCAN model on selected numeric columns and uses the output cluster label as a new feature.

  • Interactions Transformer

    The Interactions Transformer adds, divides, multiplies, and subtracts two numeric columns in the data to create a new feature. This transformation uses a smart search to identify which feature pairs to transform. Only interactions that improve the baseline model score are kept.

  • InteractionsSimple Transformer

    The InteractionsSimple Transformer adds, divides, multiplies, and subtracts two numeric columns in the data to create a new feature. This transformation randomly selects pairs of features to transform.

  • NumCatTE Transformer

    The Numeric Categorical Target Encoding Transformer calculates the mean of the response column for several selected columns. If one of the selected columns is numeric, it is first converted to categorical by binning. The mean of the response column is used as a new feature. Cross Validation is used to calculate mean response to prevent overfitting.

  • NumToCatTE Transformer

    The Numeric to Categorical Target Encoding Transformer converts numeric columns to categoricals by binning and then calculates the mean of the response column for each group. The mean of the response for the bin is used as a new feature. Cross Validation is used to calculate mean response to prevent overfitting.

  • NumToCatWoEMonotonic Transformer

    The Numeric to Categorical Weight of Evidence Monotonic Transformer converts a numeric column to categorical by binning and then calculates Weight of Evidence for each bin. The monotonic constraint ensures the bins of values are monotonically related to the Weight of Evidence value. The Weight of Evidence is used as a new feature. Weight of Evidence measures the “strength” of a grouping for separating good and bad risk and is calculated by taking the log of the ratio of distributions for a binary response column.

  • NumToCatWoE Transformer

    The Numeric to Categorical Weight of Evidence Transformer converts a numeric column to categorical by binning and then calculates Weight of Evidence for each bin. The Weight of Evidence is used as a new feature. Weight of Evidence measures the “strength” of a grouping for separating good and bad risk and is calculated by taking the log of the ratio of distributions for a binary response column.

  • Original Transformer

    The Original Transformer applies an identity transformation to a numeric column.

  • TruncSVDNum Transformer

    Truncated SVD Transformer trains a Truncated SVD model on selected numeric columns and uses the components of the truncated SVD matrix as new features.

  • TruncSVDNum cuML Transformer

    The Truncated SVD cuML Transformer runs on GPUs to train cuML accelerates Truncated SVD model on selected numeric columns and uses the components of the truncated SVD matrix as new features.

  • TSNECUMLTransformer

    Trains a cuML T-Distributed Stochastic Neighbor Embedding Model on selected numeric columns. t-SNE is a dimensionality reduction algorithm. The components of the t-SNE matrix are used as new features.

  • UMAPCUMLTransformer

    Trains a cuML Uniform Manifold Approximation and Projection Model on selected numeric columns. UMAP is a dimensionality reduction algorithm. The components of the UMAP matrix are used as new features.

  • UMAPCUMLDaskTransformer

    Trains a cuML Uniform Manifold Approximation and Projection Model on selected numeric columns. UMAP is a dimensionality reduction algorithm. The components of the UMAP matrix are used as new features

Time Series Experiments Transformers

  • DateOriginal Transformer

    The Date Original Transformer retrieves date values such as year, quarter, month, day, day of the year, week, and weekday values.

  • DateTimeOriginal Transformer

    The Date Time Original Transformer retrieves date and time values such as year, quarter, month, day, day of the year, week, weekday, hour, minute, and second values.

  • EwmaLags Transformer

    The Exponentially Weighted Moving Average (EWMA) Transformer calculates the exponentially weighted moving average of target or feature lags.

  • LagsAggregates Transformer

    The Lags Aggregates Transformer calculates aggregations of target/feature lags like mean(lag7, lag14, lag21) with support for mean, min, max, median, sum, skew, kurtosis, std. The aggregation is used as a new feature.

  • LagsInteraction Transformer

    The Lags Interaction Transformer creates target/feature lags and calculates interactions between the lags (lag2 - lag1, for instance). The interaction is used as a new feature.

  • Lags Transformer

    The Lags Transformer creates target/feature lags, possibly over groups. Each lag is used as a new feature. Lag transformers may apply to categorical (strings) features or binary/multiclass string valued targets after they have been internally numerically encoded.

  • LinearLagsRegression Transformer

    The Linear Lags Regression transformer trains a linear model on the target or feature lags to predict the current target or feature value. The linear model prediction is used as a new feature.

Categorical Transformers (String)

  • Cat Transformer

    The Cat Transformer sorts a categorical column in lexicographical order and uses the order index created as a new feature. This transformer is only enabled for models with categorical feature support. Models that have categorical feature support are LightGBM models when enable_lightgbm_cat_support is enabled or custom model recipes where _can_handle_categorical is set to True.

  • CatOriginal Transformer

    The Categorical Original Transformer applies an identity transformation that leaves categorical features as they are. This transformer works with models that can handle non-numeric feature values.

  • CVCatNumEncode Transformer

    The Cross Validation Categorical to Numeric Encoding Transformer calculates an aggregation of a numeric column for each value in a categorical column (ex: calculate the mean Temperature for each City) and uses this aggregation as a new feature.

  • CVTargetEncode Transformer (CVTE)

    The Cross Validation Target Encoding Transformer calculates the mean of the response column for each value in a categorical column and uses this as a new feature. Cross Validation is used to calculate mean response to prevent overfitting.

  • Frequent Transformer

    The Frequent Transformer calculates the frequency for each value in categorical column(s) and uses this as a new feature. This count can be either the raw count or the normalized count.

  • LexiLabelEncoder Transformer

    The Lexi Label Encoder sorts a categorical column in lexicographical order and uses the order index created as a new feature. To enable the LexiLabelEncoder transformer, set enable_lexilabel_encoding to “ON”.

  • NumCatTE Transformer

    The Numeric Categorical Target Encoding Transformer calculates the mean of the response column for several selected columns. If one of the selected columns is numeric, it is first converted to categorical by binning. The mean of the response column is used as a new feature. Cross Validation is used to calculate mean response to prevent overfitting.

  • OneHotEncoding Transformer

    The One-hot Encoding transformer converts a categorical column to a series of Boolean features by performing one-hot encoding. The Boolean features are used as new features. If there are more than a specific number of unique values in the column, then they will be binned to the max number (10 by default) in lexicographical order. This value can be changed with the ohe_bin_list config.toml configuration option.

  • SortedLE Transformer

    The Sorted Label Encoding Transformer sorts a categorical column by the response column and uses the order index created as a new feature.

  • WeightOfEvidence Transformer

    The Weight of Evidence (WoE) Transformer calculates the Weight of Evidence for each value in categorical column(s) for all possible combinations of the multiple columns. The Weight of Evidence is used as a new feature. Weight of Evidence measures the “strength” of a grouping for separating good and bad risk and is calculated by taking the log of the ratio of distributions for a binary response column.

    _images/woe.png

    This only works with a binary target variable. The likelihood needs to be created within a stratified k-fold if a fit_transform method is used. For more information, see http://ucanalytics.com/blogs/information-value-and-weight-of-evidencebanking-case/.

Text Transformers (String)

  • BERT Transformer

    The Bidirectional Encoder Representations from Transformers (BERT) Transformer creates new features for each text column based on the pre-trained model embeddings and is ideally suited for datasets that contain additional important non-text features.

Note: If your dataset is large or contains many text columns, then using the BERT transformer may significantly increase the time it takes for your experiment to complete.

  • TextBiGRU Transformer

    The Text Bidirectional GRU Transformer trains a bi-directional GRU TensorFlow model on word embeddings created from a text feature to predict the response column. The GRU prediction is used as a new feature. Cross Validation is used when training the GRU model to prevent overfitting.

  • TextCharCNN Transformer

    The Text Character CNN Transformer trains a CNN TensorFlow model on character embeddings created from a text feature to predict the response column. The CNN prediction is used as a new feature. Cross Validation is used when training the CNN model to prevent overfitting.

  • TextCNN Transformer

    The Text CNN Transformer trains a CNN TensorFlow model on word embeddings created from a text feature to predict the response column. The CNN prediction is used as a new a feature. Cross Validation is used when training the CNN model to prevent overfitting.

  • TextLinModel Transformer

    The Text Linear Model Transformer trains a linear model on a TF-IDF matrix created from a text feature to predict the response column. The linear model prediction is used as a new feature. Cross Validation is used when training the linear model to prevent overfitting.

  • Text Transformer

    The Text Transformer tokenizes a text column and creates a TFIDF matrix (term frequency-inverse document frequency) or count (count of the word) matrix. When the number of TF-IDF features exceeds the config TOML value in the list text_gene_dim_reduction_choices, dimensionality reduction is performed using truncated SVD. Selected components of the TF-IDF/Count matrix are used as new features.

  • TextOriginal Transformer

    The TextOriginal Transformer performs no feature engineering on the text column. Note that this transformer is only available for models that have text feature support. Models that have text feature support are ImageAutoModel, FTRL, BERT, and unsupervised models, in addition to custom model recipes where _can_handle_text is set to True.

Time Transformers (Date, Time)

  • Dates Transformer

    The Dates Transformer retrieves any date values, including:

    • Year

    • Quarter

    • Month

    • Day

    • Day of year

    • Week

    • Week day

    • Hour

    • Minute

    • Second

  • IsHoliday Transformer

    The Is Holiday Transformer determines if a date column is a holiday. A Boolean column indicating if the date is a holiday is added as a new feature. Creates a separate feature for holidays in the United States, United Kingdom, Germany, Mexico, and the European Central Bank. Other countries available in the python Holiday package can be added via the configuration file.

Image Transformers

  • ImageOriginal Transformer

    The Image Original Transformer passes image paths to the model without performing any feature engineering.

  • ImageVectorizer Transformer

    The Image Vectorizer Transformer uses pre-trained ImageNet models to convert a column with an image path or URI to an embeddings (vector) representation that is derived from the last global average pooling layer of the model.

    Note: Fine-tuning of the pre-trained image models can be enabled with the image-model-fine-tune expert setting.

Autoviz Recommendations Transformer

The Autoviz recommendation transformer applies the recommended transformations obtained by visualizing the dataset in Driverless AI. This transformer currently supports square_root, log and inverse operations (and their approximations using yeo-johnson power transformations for negative values).

The autoviz_recommended_transformation in the expert experiment settings list/control the transformation applied. The syntax is a dict of transformations from Autoviz {column_name: transformation} like {“DIS”:”log”,”INDUS”:”log”,”RAD”:”inverse”,”ZN”:”square_root”}. The Autoviz recommendation transformer itself can be enabled or disabled from the expert panel by included_transformers config setting.

This transformer is supported in python scoring pipelines and mojo scoring pipelines with Java Runtime (no C++ support at the moment).

Example Transformations

In this section, we will describe some of the available transformations using the example of predicting house prices on the example dataset.

Date Built

Square Footage

Num Beds

Num Baths

State

Price

01/01/1920

1700

3

2

NY

$700K

Frequent Transformer

  • the count of each categorical value in the dataset

  • the count can be either the raw count or the normalized count

Date Built

Square Footage

Num Beds

Num Baths

State

Price

Freq_State

01/01/1920

1700

3

2

NY

700,000

4,500

There are 4,500 properties in this dataset with state = NY.

Bulk Interactions Transformer

  • add, divide, multiply, and subtract two columns in the data

Date Built

Square Footage

Num Beds

Num Baths

State

Price

Interaction_NumBeds#subtract#NumBaths

01/01/1920

1700

3

2

NY

700,000

1

There is one more bedroom than there are number of bathrooms for this property.

Truncated SVD Numeric Transformer

  • truncated SVD trained on selected numeric columns of the data

  • the components of the truncated SVD will be new features

Date Built

Square Footage

Num Beds

Num Baths

State

Price

TruncSVD_Price_NumBeds_NumBaths_1

01/01/1920

1700

3

2

NY

700,000

0.632

The first component of the truncated SVD of the columns Price, Number of Beds, Number of Baths.

Dates Transformer

  • get year, get quarter, get month, get day, get day of year, get week, get week day, get hour, get minute, get second

Date Built

Square Footage

Num Beds

Num Baths

State

Price

DateBuilt_Month

01/01/1920

1700

3

2

NY

700,000

1

The home was built in the month January.

Text Transformer

  • transform text column using methods: TFIDF or count (count of the word)

  • this may be followed by dimensionality reduction using truncated SVD

Categorical Target Encoding Transformer

  • cross validation target encoding done on a categorical column

Date Built

Square Footage

Num Beds

Num Baths

State

Price

CV_TE_State

01/01/1920

1700

3

2

NY

700,000

550,000

The average price of properties in NY state is $550,000*.

*In order to prevent overfitting, Driverless AI calculates this average on out-of-fold data using cross validation.

Numeric to Categorical Target Encoding Transformer

  • numeric column converted to categorical by binning

  • cross validation target encoding done on the binned numeric column

Date Built

Square Footage

Num Beds

Num Baths

State

Price

CV_TE_SquareFootage

01/01/1920

1700

3

2

NY

700,000

345,000

The column Square Footage has been bucketed into 10 equally populated bins. This property lies in the Square Footage bucket 1,572 to 1,749. The average price of properties with this range of square footage is $345,000*.

*In order to prevent overfitting, Driverless AI calculates this average on out-of-fold data using cross validation.

Cluster Target Encoding Transformer

  • selected columns in the data are clustered

  • target encoding is done on the cluster ID

Date Built

Square Footage

Num Beds

Num Baths

State

Price

ClusterTE_4_NumBeds_NumBaths_SquareFootage

01/01/1920

1700

3

2

NY

700,000

450,000

The columns: Num Beds, Num Baths, Square Footage have been segmented into 4 clusters. The average price of properties in the same cluster as the selected property is $450,000*.

*In order to prevent overfitting, Driverless AI calculates this average on out-of-fold data using cross validation.

Cluster Distance Transformer

  • selected columns in the data are clustered

  • the distance to a chosen cluster center is calculated

Date Built

Square Footage

Num Beds

Num Baths

State

Price

ClusterDist_4_NumBeds_NumBaths_SquareFootage_1

01/01/1920

1700

3

2

NY

700,000

0.83

The columns: Num Beds, Num Baths, Square Footage have been segmented into 4 clusters. The difference from this record to Cluster 1 is 0.83.