CartLearner
CartLearner ¶
CartLearner(label: str, task: Task = CLASSIFICATION, *, weights: Optional[str] = None, ranking_group: Optional[str] = None, uplift_treatment: Optional[str] = None, features: Optional[ColumnDefs] = None, include_all_columns: bool = False, max_vocab_count: int = 2000, min_vocab_frequency: int = 5, discretize_numerical_columns: bool = False, num_discretized_numerical_bins: int = 255, max_num_scanned_rows_to_infer_semantic: int = 100000, max_num_scanned_rows_to_compute_statistics: int = 100000, data_spec: Optional[DataSpecification] = None, allow_na_conditions: bool = False, categorical_algorithm: str = 'CART', categorical_set_split_greedy_sampling: float = 0.1, categorical_set_split_max_num_items: int = -1, categorical_set_split_min_item_frequency: int = 1, growing_strategy: str = 'LOCAL', honest: bool = False, honest_fixed_separation: bool = False, honest_ratio_leaf_examples: float = 0.5, in_split_min_examples_check: bool = True, keep_non_leaf_label_distribution: bool = True, max_depth: int = 16, max_num_nodes: Optional[int] = None, maximum_model_size_in_memory_in_bytes: float = -1.0, maximum_training_duration_seconds: float = -1.0, mhld_oblique_max_num_attributes: Optional[int] = None, mhld_oblique_sample_attributes: Optional[bool] = None, min_examples: int = 5, missing_value_policy: str = 'GLOBAL_IMPUTATION', num_candidate_attributes: Optional[int] = -1, num_candidate_attributes_ratio: Optional[float] = None, pure_serving_model: bool = False, random_seed: int = 123456, sorting_strategy: str = 'IN_NODE', sparse_oblique_max_num_projections: Optional[int] = None, sparse_oblique_normalization: Optional[str] = None, sparse_oblique_num_projections_exponent: Optional[float] = None, sparse_oblique_projection_density_factor: Optional[float] = None, sparse_oblique_weights: Optional[str] = None, split_axis: str = 'AXIS_ALIGNED', uplift_min_examples_in_treatment: int = 5, uplift_split_score: str = 'KULLBACK_LEIBLER', validation_ratio: float = 0.1, working_dir: Optional[str] = None, num_threads: Optional[int] = None, tuner: Optional[AbstractTuner] = None, explicit_args: Optional[Set[str]] = None)
Bases: GenericLearner
Cart learning algorithm.
A CART (Classification and Regression Trees) a decision tree. The non-leaf nodes contains conditions (also known as splits) while the leaf nodes contain prediction values. The training dataset is divided in two parts. The first is used to grow the tree while the second is used to prune the tree.
Usage example:
import ydf
import pandas as pd
dataset = pd.read_csv("project/dataset.csv")
model = ydf.CartLearner().train(dataset)
print(model.describe())
Hyperparameters are configured to give reasonable results for typical
datasets. Hyperparameters can also be modified manually (see descriptions)
below or by applying the hyperparameter templates available with
CartLearner.hyperparameter_templates()
(see this function's documentation for
details).
Attributes:
label: Label of the dataset. The label column
should not be identified as a feature in the features
parameter.
task: Task to solve (e.g. Task.CLASSIFICATION, Task.REGRESSION,
Task.RANKING, Task.CATEGORICAL_UPLIFT, Task.NUMERICAL_UPLIFT).
weights: Name of a feature that identifies the weight of each example. If
weights are not specified, unit weights are assumed. The weight column
should not be identified as a feature in the features
parameter.
ranking_group: Only for task=Task.RANKING
. Name of a feature
that identifies queries in a query/document ranking task. The ranking
group should not be identified as a feature in the features
parameter.
uplift_treatment: Only for task=Task.CATEGORICAL_UPLIFT
and task=Task
.
NUMERICAL_UPLIFT. Name of a numerical feature that identifies the
treatment in an uplift problem. The value 0 is reserved for the control
treatment. Currently, only 0/1 binary treatments are supported.
features: If None, all columns are used as features. The semantic of the
features is determined automatically. Otherwise, if
include_all_columns=False (default) only the column listed in features
are imported. If include_all_columns=True, all the columns are imported as
features and only the semantic of the columns NOT in columns
is
determined automatically. If specified, defines the order of the features
- any non-listed features are appended in-order after the specified
features (if include_all_columns=True).
The label, weights, uplift treatment and ranking_group columns should not
be specified as features.
include_all_columns: See features
.
max_vocab_count: Maximum size of the vocabulary of CATEGORICAL and
CATEGORICAL_SET columns stored as strings. If more unique values exist,
only the most frequent values are kept, and the remaining values are
considered as out-of-vocabulary.
min_vocab_frequency: Minimum number of occurrence of a value for CATEGORICAL
and CATEGORICAL_SET columns. Value observed less than
min_vocab_frequency
are considered as out-of-vocabulary.
discretize_numerical_columns: If true, discretize all the numerical columns
before training. Discretized numerical columns are faster to train with,
but they can have a negative impact on the model quality. Using
discretize_numerical_columns=True
is equivalent as setting the column
semantic DISCRETIZED_NUMERICAL in the column
argument. See the
definition of DISCRETIZED_NUMERICAL for more details.
num_discretized_numerical_bins: Number of bins used when disretizing
numerical columns.
max_num_scanned_rows_to_infer_semantic: Number of rows to scan when
inferring the column's semantic if it is not explicitly specified. Only
used when reading from file, in-memory datasets are always read in full.
Setting this to a lower number will speed up dataset reading, but might
result in incorrect column semantics. Set to -1 to scan the entire
dataset.
max_num_scanned_rows_to_compute_statistics: Number of rows to scan when
computing a column's statistics. Only used when reading from file,
in-memory datasets are always read in full. A column's statistics include
the dictionary for categorical features and the mean / min / max for
numerical features. Setting this to a lower number will speed up dataset
reading, but skew statistics in the dataspec, which can hurt model quality
(e.g. if an important category of a categorical feature is considered
OOV). Set to -1 to scan the entire dataset.
data_spec: Dataspec to be used (advanced). If a data spec is given,
columns
, include_all_columns
, max_vocab_count
,
min_vocab_frequency
, discretize_numerical_columns
and
num_discretized_numerical_bins
will be ignored.
allow_na_conditions: If true, the tree training evaluates conditions of the
type X is NA
i.e. X is missing
. Default: False.
categorical_algorithm: How to learn splits on categorical attributes.
- CART
: CART algorithm. Find categorical splits of the form "value \in
mask". The solution is exact for binary classification, regression and
ranking. It is approximated for multi-class classification. This is a
good first algorithm to use. In case of overfitting (very small
dataset, large dictionary), the "random" algorithm is a good
alternative.
- ONE_HOT
: One-hot encoding. Find the optimal categorical split of the
form "attribute == param". This method is similar (but more efficient)
than converting converting each possible categorical value into a
boolean feature. This method is available for comparison purpose and
generally performs worse than other alternatives.
- RANDOM
: Best splits among a set of random candidate. Find the a
categorical split of the form "value \in mask" using a random search.
This solution can be seen as an approximation of the CART algorithm.
This method is a strong alternative to CART. This algorithm is inspired
from section "5.1 Categorical Variables" of "Random Forest", 2001.
Default: "CART".
categorical_set_split_greedy_sampling: For categorical set splits e.g.
texts. Probability for a categorical value to be a candidate for the
positive set. The sampling is applied once per node (i.e. not at every
step of the greedy optimization). Default: 0.1.
categorical_set_split_max_num_items: For categorical set splits e.g. texts.
Maximum number of items (prior to the sampling). If more items are
available, the least frequent items are ignored. Changing this value is
similar to change the "max_vocab_count" before loading the dataset, with
the following exception: With max_vocab_count
, all the remaining items
are grouped in a special Out-of-vocabulary item. With max_num_items
,
this is not the case. Default: -1.
categorical_set_split_min_item_frequency: For categorical set splits e.g.
texts. Minimum number of occurrences of an item to be considered.
Default: 1.
growing_strategy: How to grow the tree.
- LOCAL
: Each node is split independently of the other nodes. In other
words, as long as a node satisfy the splits "constraints (e.g. maximum
depth, minimum number of observations), the node will be split. This is
the "classical" way to grow decision trees.
- BEST_FIRST_GLOBAL
: The node with the best loss reduction among all
the nodes of the tree is selected for splitting. This method is also
called "best first" or "leaf-wise growth". See "Best-first decision
tree learning", Shi and "Additive logistic regression : A statistical
view of boosting", Friedman for more details. Default: "LOCAL".
honest: In honest trees, different training examples are used to infer the
structure and the leaf values. This regularization technique trades
examples for bias estimates. It might increase or reduce the quality of
the model. See "Generalized Random Forests", Athey et al. In this paper,
Honest trees are trained with the Random Forest algorithm with a sampling
without replacement. Default: False.
honest_fixed_separation: For honest trees only i.e. honest=true. If true, a
new random separation is generated for each tree. If false, the same
separation is used for all the trees (e.g., in Gradient Boosted Trees
containing multiple trees). Default: False.
honest_ratio_leaf_examples: For honest trees only i.e. honest=true. Ratio
of examples used to set the leaf values. Default: 0.5.
in_split_min_examples_check: Whether to check the min_examples
constraint
in the split search (i.e. splits leading to one child having less than
min_examples
examples are considered invalid) or before the split
search (i.e. a node can be derived only if it contains more than
min_examples
examples). If false, there can be nodes with less than
min_examples
training examples. Default: True.
keep_non_leaf_label_distribution: Whether to keep the node value (i.e. the
distribution of the labels of the training examples) of non-leaf nodes.
This information is not used during serving, however it can be used for
model interpretation as well as hyper parameter tuning. This can take
lots of space, sometimes accounting for half of the model size. Default:
True.
max_depth: Maximum depth of the tree. max_depth=1
means that all trees
will be roots. max_depth=-1
means that tree depth is not restricted by
this parameter. Values <= -2 will be ignored. Default: 16.
max_num_nodes: Maximum number of nodes in the tree. Set to -1 to disable
this limit. Only available for growing_strategy=BEST_FIRST_GLOBAL
.
Default: None.
maximum_model_size_in_memory_in_bytes: Limit the size of the model when
stored in ram. Different algorithms can enforce this limit differently.
Note that when models are compiled into an inference, the size of the
inference engine is generally much smaller than the original model.
Default: -1.0.
maximum_training_duration_seconds: Maximum training duration of the model
expressed in seconds. Each learning algorithm is free to use this
parameter at it sees fit. Enabling maximum training duration makes the
model training non-deterministic. Default: -1.0.
mhld_oblique_max_num_attributes: For MHLD oblique splits i.e.
split_axis=MHLD_OBLIQUE
. Maximum number of attributes in the
projection. Increasing this value increases the training time. Decreasing
this value acts as a regularization. The value should be in [2,
num_numerical_features]. If the value is above the total number of
numerical features, the value is capped automatically. The value 1 is
allowed but results in ordinary (non-oblique) splits. Default: None.
mhld_oblique_sample_attributes: For MHLD oblique splits i.e.
split_axis=MHLD_OBLIQUE
. If true, applies the attribute sampling
controlled by the "num_candidate_attributes" or
"num_candidate_attributes_ratio" parameters. If false, all the attributes
are tested. Default: None.
min_examples: Minimum number of examples in a node. Default: 5.
missing_value_policy: Method used to handle missing attribute values.
- GLOBAL_IMPUTATION
: Missing attribute values are imputed, with the
mean (in case of numerical attribute) or the most-frequent-item (in
case of categorical attribute) computed on the entire dataset (i.e. the
information contained in the data spec).
- LOCAL_IMPUTATION
: Missing attribute values are imputed with the mean
(numerical attribute) or most-frequent-item (in the case of categorical
attribute) evaluated on the training examples in the current node.
- RANDOM_LOCAL_IMPUTATION
: Missing attribute values are imputed from
randomly sampled values from the training examples in the current node.
This method was proposed by Clinic et al. in "Random Survival Forests"
(https://projecteuclid.org/download/pdfview_1/euclid.aoas/1223908043).
Default: "GLOBAL_IMPUTATION".
num_candidate_attributes: Number of unique valid attributes tested for each
node. An attribute is valid if it has at least a valid split. If
num_candidate_attributes=0
, the value is set to the classical default
value for Random Forest: sqrt(number of input attributes)
in case of
classification and number_of_input_attributes / 3
in case of
regression. If num_candidate_attributes=-1
, all the attributes are
tested. Default: -1.
num_candidate_attributes_ratio: Ratio of attributes tested at each node. If
set, it is equivalent to num_candidate_attributes =
number_of_input_features x num_candidate_attributes_ratio
. The possible
values are between ]0, and 1] as well as -1. If not set or equal to -1,
the num_candidate_attributes
is used. Default: None.
pure_serving_model: Clear the model from any information that is not
required for model serving. This includes debugging, model interpretation
and other meta-data. The size of the serialized model can be reduced
significatively (50% model size reduction is common). This parameter has
no impact on the quality, serving speed or RAM usage of model serving.
Default: False.
random_seed: Random seed for the training of the model. Learners are
expected to be deterministic by the random seed. Default: 123456.
sorting_strategy: How are sorted the numerical features in order to find
the splits
- AUTO: Selects the most efficient method among IN_NODE, FORCE_PRESORT,
and LAYER.
- IN_NODE: The features are sorted just before being used in the node.
This solution is slow but consumes little amount of memory.
- FORCE_PRESORT: The features are pre-sorted at the start of the
training. This solution is faster but consumes much more memory than
IN_NODE.
- PRESORT: Automatically choose between FORCE_PRESORT and IN_NODE.
. Default: "IN_NODE".
sparse_oblique_max_num_projections: For sparse oblique splits i.e.
split_axis=SPARSE_OBLIQUE
. Maximum number of projections (applied after
the num_projections_exponent).
Oblique splits try out max(p^num_projections_exponent,
max_num_projections) random projections for choosing a split, where p is
the number of numerical features. Increasing "max_num_projections"
increases the training time but not the inference time. In late stage
model development, if every bit of accuracy if important, increase this
value.
The paper "Sparse Projection Oblique Random Forests" (Tomita et al, 2020)
does not define this hyperparameter. Default: None.
sparse_oblique_normalization: For sparse oblique splits i.e.
split_axis=SPARSE_OBLIQUE
. Normalization applied on the features,
before applying the sparse oblique projections.
- NONE
: No normalization.
- STANDARD_DEVIATION
: Normalize the feature by the estimated standard
deviation on the entire train dataset. Also known as Z-Score
normalization.
- MIN_MAX
: Normalize the feature by the range (i.e. max-min) estimated
on the entire train dataset. Default: None.
sparse_oblique_num_projections_exponent: For sparse oblique splits i.e.
split_axis=SPARSE_OBLIQUE
. Controls of the number of random projections
to test at each node.
Increasing this value very likely improves the quality of the model,
drastically increases the training time, and doe not impact the inference
time.
Oblique splits try out max(p^num_projections_exponent,
max_num_projections) random projections for choosing a split, where p is
the number of numerical features. Therefore, increasing this
num_projections_exponent
and possibly max_num_projections
may improve
model quality, but will also significantly increase training time.
Note that the complexity of (classic) Random Forests is roughly
proportional to num_projections_exponent=0.5
, since it considers
sqrt(num_features) for a split. The complexity of (classic) GBDT is
roughly proportional to num_projections_exponent=1
, since it considers
all features for a split.
The paper "Sparse Projection Oblique Random Forests" (Tomita et al, 2020)
recommends values in [1/4, 2]. Default: None.
sparse_oblique_projection_density_factor: Density of the projections as an
exponent of the number of features. Independently for each projection,
each feature has a probability "projection_density_factor / num_features"
to be considered in the projection.
The paper "Sparse Projection Oblique Random Forests" (Tomita et al, 2020)
calls this parameter lambda
and recommends values in [1, 5].
Increasing this value increases training and inference time (on average).
This value is best tuned for each dataset. Default: None.
sparse_oblique_weights: For sparse oblique splits i.e.
split_axis=SPARSE_OBLIQUE
. Possible values:
- BINARY
: The oblique weights are sampled in {-1,1} (default).
- CONTINUOUS
: The oblique weights are be sampled in [-1,1]. Default:
None.
split_axis: What structure of split to consider for numerical features.
- AXIS_ALIGNED
: Axis aligned splits (i.e. one condition at a time).
This is the "classical" way to train a tree. Default value.
- SPARSE_OBLIQUE
: Sparse oblique splits (i.e. random splits on a small
number of features) from "Sparse Projection Oblique Random Forests",
Tomita et al., 2020.
- MHLD_OBLIQUE
: Multi-class Hellinger Linear Discriminant splits from
"Classification Based on Multivariate Contrast Patterns",
Canete-Sifuentes et al., 2029 Default: "AXIS_ALIGNED".
uplift_min_examples_in_treatment: For uplift models only. Minimum number of
examples per treatment in a node. Default: 5.
uplift_split_score: For uplift models only. Splitter score i.e. score
optimized by the splitters. The scores are introduced in "Decision trees
for uplift modeling with single and multiple treatments", Rzepakowski et
al. Notation: p
probability / average value of the positive outcome,
q
probability / average value in the control group.
- KULLBACK_LEIBLER
or KL
: - p log (p/q)
- EUCLIDEAN_DISTANCE
or ED
: (p-q)^2
- CHI_SQUARED
or CS
: (p-q)^2/q
Default: "KULLBACK_LEIBLER".
validation_ratio: Ratio of the training dataset used to create the
validation dataset for pruning the tree. If set to 0, the entire dataset
is used for training, and the tree is not pruned. Default: 0.1.
working_dir: Path to a directory available for the learning algorithm to
store intermediate computation results. Depending on the learning
algorithm and parameters, the working_dir might be optional, required, or
ignored. For instance, distributed training algorithm always need a
"working_dir", and the gradient boosted tree and hyper-parameter tuners
will export artefacts to the "working_dir" if provided.
num_threads: Number of threads used to train the model. Different learning
algorithms use multi-threading differently and with different degree of
efficiency. If None
, num_threads
will be automatically set to the
number of processors (up to a maximum of 32; or set to 6 if the number of
processors is not available). Making num_threads
significantly larger
than the number of processors can slow-down the training speed. The
default value logic might change in the future.
tuner: If set, automatically select the best hyperparameters using the
provided tuner. When using distributed training, the tuning is
distributed.
explicit_args: Helper argument for internal use. Throws if supplied
explicitly by the user.
hyperparameters
property
¶
A (mutable) dictionary of this learner's hyperparameters.
This object can be used to inspect or modify hyperparameters after creating
the learner. Modifying hyperparameters after constructing the learner is
suitable for some advanced use cases. Since this approach bypasses some
feasibility checks for the given set of hyperparameters, it generally better
to re-create the learner for each model. The current set of hyperparameters
can be validated manually with validate_hyperparameters()
.
cross_validation ¶
cross_validation(ds: InputDataset, folds: int = 10, bootstrapping: Union[bool, int] = False, parallel_evaluations: int = 1) -> Evaluation
Cross-validates the learner and return the evaluation.
Usage example:
import pandas as pd
import ydf
dataset = pd.read_csv("my_dataset.csv")
learner = ydf.RandomForestLearner(label="label")
evaluation = learner.cross_validation(dataset)
# In a notebook, display an interractive evaluation
evaluation
# Print the evaluation
print(evaluation)
# Look at specific metrics
print(evaluation.accuracy)
Parameters:
Name | Type | Description | Default |
---|---|---|---|
ds
|
InputDataset
|
Dataset for the cross-validation. |
required |
folds
|
int
|
Number of cross-validation folds. |
10
|
bootstrapping
|
Union[bool, int]
|
Controls whether bootstrapping is used to evaluate the confidence intervals and statistical tests (i.e., all the metrics ending with "[B]"). If set to false, bootstrapping is disabled. If set to true, bootstrapping is enabled and 2000 bootstrapping samples are used. If set to an integer, it specifies the number of bootstrapping samples to use. In this case, if the number is less than 100, an error is raised as bootstrapping will not yield useful results. |
False
|
parallel_evaluations
|
int
|
Number of model to train and evaluate in parallel
using multi-threading. Note that each model is potentially already
trained with multithreading (see |
1
|
Returns:
Type | Description |
---|---|
Evaluation
|
The cross-validation evaluation. |
hyperparameter_templates
classmethod
¶
train ¶
train(ds: InputDataset, valid: Optional[InputDataset] = None, verbose: Optional[Union[int, bool]] = None) -> RandomForestModel
Trains a model on the given dataset.
Options for dataset reading are given on the learner. Consult the documentation of the learner or ydf.create_vertical_dataset() for additional information on dataset reading in YDF.
Usage example:
import ydf
import pandas as pd
train_ds = pd.read_csv(...)
learner = ydf.CartLearner(label="label")
model = learner.train(train_ds)
print(model.summary())
If training is interrupted (for example, by interrupting the cell execution in Colab), the model will be returned to the state it was in at the moment of interruption.
Parameters:
Name | Type | Description | Default |
---|---|---|---|
ds
|
InputDataset
|
Training dataset. |
required |
valid
|
Optional[InputDataset]
|
Optional validation dataset. Some learners, such as Random Forest, do not need validation dataset. Some learners, such as GradientBoostedTrees, automatically extract a validation dataset from the training dataset if the validation dataset is not provided. |
None
|
verbose
|
Optional[Union[int, bool]]
|
Verbose level during training. If None, uses the global verbose
level of |
None
|
Returns:
Type | Description |
---|---|
RandomForestModel
|
A trained model. |
validate_hyperparameters ¶
Returns None if the hyperparameters are valid, raises otherwise.
This method is called automatically before training, but users may call it to fail early. It makes sense to call this method when changing manually the hyper-paramters of the learner. This is a relatively advanced approach that is not recommende (it is better to re-create the learner in most cases).
Usage example: