8000 feat: support data split for evaluation in linear and ensemble models by shobsi · Pull Request #1081 · googleapis/python-bigquery-dataframes · GitHub
[go: up one dir, main page]
Skip to content

feat: support data split for evaluation in linear and ensemble models #1081

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
shobsi wants to merge 8 commits into from
Closed

feat: support data split for evaluation in linear and ensemble models #1081

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
c7a41ff
feat: support data split method in `LinearRegression`
shobsi Oct 14, 2024
98b9f85
place new params after the existing ones
shobsi Oct 14, 2024
088d535
Merge remote-tracking branch 'refs/remotes/github/main' into shobs-ml…
shobsi Oct 14, 2024
70e17a5
fix unit test to expect no_split instead of NO_SPLIT
shobsi Oct 14, 2024
5211186
update documentation for LinearRegression
shobsi Oct 14, 2024< 8000 /relative-time>
32a6f71
support data split in LogisticRegression model
shobsi Oct 15, 2024
542d593
support data split in xgb and random forest models
shobsi Oct 15, 2024
e37869d
Merge remote-tracking branch 'refs/remotes/github/main' into shobs-ml…
shobsi Oct 15, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
105 changes: 92 additions & 13 deletions bigframes/ml/ensemble.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -17,7 +17,7 @@

from __future__ import annotations

from typing import Dict, List, Literal, Optional, Union
from typing import List, Literal, Optional, Union

import bigframes_vendored.sklearn.ensemble._forest
import bigframes_vendored.xgboost.sklearn
Expand Down Expand Up @@ -47,6 +47,9 @@
"max_iterations": "maxIterations",
"enable_global_explain": "enableGlobalExplain",
"xgboost_version": "xgboostVersion",
"data_split_method": "dataSplitMethod",
"data_split_eval_fraction": "dataSplitEvalFraction",
"data_split_col": "dataSplitColumn",
}


Expand Down Expand Up @@ -78,6 +81,15 @@ def __init__(
tol: float = 0.01,
enable_global_explain: bool = False,
xgboost_version: Literal["0.9", "1.1"] = "0.9",
data_split_method: Literal[
"auto_split",
"random",
"custom",
"seq",
"no_split",
] = "no_split",
data_split_eval_fraction: Optional[float] = None,
data_split_col: Optional[str] = None,
):
self.n_estimators = n_estimators
self.booster = booster
Expand All @@ -97,6 +109,9 @@ def __init__(
self.tol = tol
self.enable_global_explain = enable_global_explain
self.xgboost_version = xgboost_version
self.data_split_method = data_split_method
self.data_split_eval_fraction = data_split_eval_fraction
self.data_split_col = data_split_col
self._bqml_model: Optional[core.BqmlModel] = None
self._bqml_model_factory = globals.bqml_model_factory()

Expand All @@ -115,11 +130,11 @@ def _from_bq(
return model

@property
def _bqml_options(self) -> Dict[str, str | int | bool | float | List[str]]:
def _bqml_options(self) -> dict:
"""The model options as they will be set for BQML"""
return {
options = {
"model_type": "BOOSTED_TREE_REGRESSOR",
"data_split_method": "NO_SPLIT",
"data_split_method": self.data_split_method,
"early_stop": True,
"num_parallel_tree": self.n_estimators,
"booster_type": self.booster,
Expand All @@ -140,6 +155,13 @@ def _bqml_options(self) -> Dict[str, str | int | bool | float | List[str]]:
"xgboost_version": self.xgboost_version,
}

if self.data_split_eval_fraction is not None:
options["data_split_eval_fraction"] = self.data_split_eval_fraction
if self.data_split_col is not None:
options["data_split_col"] = self.data_split_col

return options

def _fit(
self,
X: Union[bpd.DataFrame, bpd.Series],
Expand Down Expand Up @@ -227,6 +249,15 @@ def __init__(
tol: float = 0.01,
enable_global_explain: bool = False,
xgboost_version: Literal["0.9", "1.1"] = "0.9",
data_split_method: Literal[
"auto_split",
"random",
"custom",
"seq",
"no_split",
] = "no_split",
data_split_eval_fraction: Optional[float] = None,
data_split_col: Optional[str] = None,
):
self.n_estimators = n_estimators
self.booster = booster
Expand All @@ -246,6 +277,9 @@ def __init__(
self.tol = tol
self.enable_global_explain = enable_global_explain
self.xgboost_version = xgboost_version
self.data_split_method = data_split_method
self.data_split_eval_fraction = data_split_eval_fraction
self.data_split_col = data_split_col
self._bqml_model: Optional[core.BqmlModel] = None
self._bqml_model_factory = globals.bqml_model_factory()

Expand All @@ -264,11 +298,11 @@ def _from_bq(
return model

@property
def _bqml_options(self) -> Dict[str, str | int | bool | float | List[str]]:
def _bqml_options(self) -> dict:
"""The model options as they will be set for BQML"""
return {
options = {
"model_type": "BOOSTED_TREE_CLASSIFIER",
"data_split_method": "NO_SPLIT",
"data_split_method": self.data_split_method,
"early_stop": True,
"num_parallel_tree": self.n_estimators,
"booster_type": self.booster,
Expand All @@ -289,6 +323,13 @@ def _bqml_options(self) -> Dict[str, str | int | bool | float | List[str]]:
"xgboost_version": self.xgboost_version,
}

if self.data_split_eval_fraction is not None:
options["data_split_eval_fraction"] = self.data_split_eval_fraction
if self.data_split_col is not None:
options["data_split_col"] = self.data_split_col

return options

def _fit(
self,
X: Union[bpd.DataFrame, bpd.Series],
Expand Down Expand Up @@ -370,6 +411,15 @@ def __init__(
tol: float = 0.01,
enable_global_explain: bool = False,
xgboost_version: Literal["0.9", "1.1"] = "0.9",
data_split_method: Literal[
"auto_split",
"random",
"custom",
"seq",
"no_split",
] = "no_split",
data_split_eval_fraction: Optional[float] = None,
data_split_col: Optional[str] = None,
):
self.n_estimators = n_estimators
self.tree_method = tree_method
Expand All @@ -385,6 +435,9 @@ def __init__(
self.tol = tol
self.enable_global_explain = enable_global_explain
self.xgboost_version = xgboost_version
self.data_split_method = data_split_method
self.data_split_eval_fraction = data_split_eval_fraction
self.data_split_col = data_split_col
self._bqml_model: Optional[core.BqmlModel] = None
self._bqml_model_factory = globals.bqml_model_factory()

Expand All @@ -403,9 +456,9 @@ def _from_bq(
return model

@property
def _bqml_options(self) -> Dict[str, str | int | bool | float | List[str]]:
def _bqml_options(self) -> dict:
"""The model options as they will be set for BQML"""
return {
options = {
"model_type": "RANDOM_FOREST_REGRESSOR",
"early_stop": True,
"num_parallel_tree": self.n_estimators,
Expand All @@ -420,11 +473,18 @@ def _bqml_options(self) -> Dict[str, str | int | bool | float | List[str]]:
"l1_reg": self.reg_alpha,
"l2_reg": self.reg_lambda,
"min_rel_progress": self.tol,
"data_split_method": "NO_SPLIT",
"data_split_method": self.data_split_method,
"enable_global_explain": self.enable_global_explain,
"xgboost_version": self.xgboost_version,
}

if self.data_split_eval_fraction is not None:
options["data_split_eval_fraction"] = self.data_split_eval_fraction
if self.data_split_col is not None:
options["data_split_col"] = self.data_split_col

return options

def _fit(
self,
X: Union[bpd.DataFrame, bpd.Series],
Expand Down Expand Up @@ -526,6 +586,15 @@ def __init__(
tol: float = 0.01,
enable_global_explain: bool = False,
xgboost_version: Literal["0.9", "1.1"] = "0.9",
data_split_method: Literal[
"auto_split",
"random",
"custom",
"seq",
"no_split",
] = "no_split",
data_split_eval_fraction: Optional[float] = None,
data_split_col: Optional[str] = None,
):
self.n_estimators = n_estimators
self.tree_method = tree_method
Expand All @@ -541,6 +610,9 @@ def __init__(
self.tol = tol
self.enable_global_explain = enable_global_explain
self.xgboost_version = xgboost_version
self.data_split_method = data_split_method
self.data_split_eval_fraction = data_split_eval_fraction
self.data_split_col = data_split_col
self._bqml_model: Optional[core.BqmlModel] = None
self._bqml_model_factory = globals.bqml_model_factory()

Expand All @@ -559,9 +631,9 @@ def _from_bq(
return model

@property
def _bqml_options(self) -> Dict[str, str | int | bool | float | List[str]]:
def _bqml_options(self) -> dict:
"""The model options as they will be set for BQML"""
return {
options = {
"model_type": "RANDOM_FOREST_CLASSIFIER",
"early_stop": True,
"num_parallel_tree": self.n_estimators,
Expand All @@ -576,11 +648,18 @@ def _bqml_options(self) -> Dict[str, str | int | bool | float | List[str]]:
"l1_reg": self.reg_alpha,
"l2_reg": self.reg_lambda,
"min_rel_progress": self.tol,
"data_split_method": "NO_SPLIT",
"data_split_method": self.data_split_method,
"enable_global_explain": self.enable_global_explain,
"xgboost_version": self.xgboost_version,
}

if self.data_split_eval_fraction is not None:
options["data_split_eval_fraction"] = self.data_split_eval_fraction
if self.data_split_col is not None:
options["data_split_col"] = self.data_split_col

return options

def _fit(
self,
X: Union[bpd.DataFrame, bpd.Series],
Expand Down
39 changes: 37 additions & 2 deletions bigframes/ml/linear_model.py
View file
Open in desktop
< 61F6 td class="blob-code blob-code-addition js-file-line"> "auto_split",
Original file line number Diff line number Diff line change
Expand Up @@ -42,6 +42,9 @@
"warm_start": "warmStart",
"calculate_p_values": "calculatePValues",
"enable_global_explain": "enableGlobalExplain",
"data_split_method": "dataSplitMethod",
"data_split_eval_fraction": "dataSplitEvalFraction",
"data_split_col": "dataSplitColumn",
}


Expand Down Expand Up @@ -69,6 +72,15 @@ def __init__(
ls_init_learning_rate: Optional[float] = None,
calculate_p_values: bool = False,
enable_global_explain: bool = False,
data_split_method: Literal[
"random",
"custom",
"seq",
"no_split",
] = "no_split",
data_split_eval_fraction: Optional[float] = None,
data_split_col: Optional[str] = None,
):
self.optimize_strategy = optimize_strategy
self.fit_intercept = fit_intercept
Expand All @@ -82,6 +94,9 @@ def __init__(
self.ls_init_learning_rate = ls_init_learning_rate
self.calculate_p_values = calculate_p_values
self.enable_global_explain = enable_global_explain
self.data_split_method = data_split_method
self.data_split_eval_fraction = data_split_eval_fraction
self.data_split_col = data_split_col
self._bqml_model: Optional[core.BqmlModel] = None
self._bqml_model_factory = globals.bqml_model_factory()

Expand All @@ -104,7 +119,7 @@ def _bqml_options(self) -> dict:
"""The model options as they will be set for BQML"""
options = {
"model_type": "LINEAR_REG",
"data_split_method": "NO_SPLIT",
"data_split_method": self.data_split_method,
"optimize_strategy": self.optimize_strategy,
"fit_intercept": self.fit_intercept,
"l2_reg": self.l2_reg,
Expand All @@ -123,6 +138,10 @@ def _bqml_options(self) -> dict:
# Even presenting warm_start returns error for NORMAL_EQUATION optimizer
if self.warm_start:
options["warm_start"] = self.warm_start
if self.data_split_eval_fraction is not None:
options["data_split_eval_fraction"] = self.data_split_eval_fraction
if self.data_split_col is not None:
options["data_split_col"] = self.data_split_col

return options

Expand Down Expand Up @@ -209,6 +228,15 @@ def __init__(
calculate_p_values: bool = False,
enable_global_explain: bool = False,
class_weight: Optional[Union[Literal["balanced"], Dict[str, float]]] = None,
data_split_method: Literal[
"auto_split",
"random",
"custom",
"seq",
"no_split",
] = "no_split",
data_split_eval_fraction: Optional[float] = None,
data_split_col: Optional[str] = None,
):
self.optimize_strategy = optimize_strategy
self.fit_intercept = fit_intercept
Expand All @@ -223,6 +251,9 @@ def __init__(
self.calculate_p_values = calculate_p_values
self.enable_global_explain = enable_global_explain
self.class_weight = class_weight
self.data_split_method = data_split_method
self.data_split_eval_fraction = data_split_eval_fraction
self.data_split_col = data_split_col
self._auto_class_weight = class_weight == "balanced"
self._bqml_model: Optional[core.BqmlModel] = None
self._bqml_model_factory = globals.bqml_model_factory()
Expand Down Expand Up @@ -253,7 +284,7 @@ def _bqml_options(self) -> dict:
"""The model options as they will be set for BQML"""
options = {
"model_type": "LOGISTIC_REG",
"data_split_method": "NO_SPLIT",
"data_split_method": self.data_split_method,
"fit_intercept": self.fit_intercept,
"auto_class_weights": self._auto_class_weight,
"optimize_strategy": self.optimize_strategy,
Expand All @@ -275,6 +306,10 @@ def _bqml_options(self) -> dict:
# Even presenting warm_start returns error for NORMAL_EQUATION optimizer
if self.warm_start:
options["warm_start"] = self.warm_start
if self.data_split_eval_fraction is not None:
options["data_split_eval_fraction"] = self.data_split_eval_fraction
if self.data_split_col is not None:
options["data_split_col"] = self.data_split_col

return options

Expand Down
Loading
0