logml.analysis.items.modeling

ML Analysis and related routines

Functions

`get_stratum`(global_cfg, stratum_id)	Returns a corresponding Strata for a given stratum id.
`stratum_high_resources_estimator`(res[, ...])	Default estimation of high resources requirements based on dataset rows and columns number.

Classes

`AggregateFeatureImportanceStep`(args, *kwargs)	FI Aggregation Step.
`EdaStep`(args, *kwargs)	EDA analysis
`ExtractFeatureImportanceStep`(args, *kwargs)	FI extraction step.
`GenerateReportStep`(args, *kwargs)	Generate Report Step.
`ModelingAnalysisResult`([selected_models, ...])	Modeling analysis result.
`ModelingItemBase`(args, *kwargs)	Wraps analysis algo into data loading/saving and config.
`ModelingTransformer`(args, *kwargs)	Wraps analysis algo into data loading/saving and config.
`PackageReportStep`(args, *kwargs)	Package report step
`SelectModelsStep`(args, *kwargs)	Models selection step.
`SurvivalAnalysisResult`([stratum_id, ...])	Survival step result
`SurvivalAnalysisStep`(args, *kwargs)	Survival analysis step.
`TrainModelStep`(args, *kwargs)	Train Model Step.

logml.analysis.items.modeling.get_stratum(global_cfg, stratum_id: str) → Optional[logml.configuration.stratification.Strata]: Returns a corresponding Strata for a given stratum id.

class logml.analysis.items.modeling.ModelingConfigBase

Bases: pydantic.main.BaseModel

Config with common modeling params

Show JSON schema

{
   "title": "ModelingConfigBase",
   "description": "Config with common modeling params",
   "type": "object",
   "properties": {
      "stratum_id": {
         "title": "Stratum Id",
         "type": "string"
      },
      "problem_id": {
         "title": "Problem Id",
         "type": "string"
      }
   },
   "required": [
      "stratum_id"
   ]
}

Fields

problem_id (Optional[str])
stratum_id (str)

field stratum_id: str [Required]

field problem_id: Optional[str] [Required]

class logml.analysis.items.modeling.ModelingItemBase(*args, **kwargs)

Bases: logml.analysis.base_item.AnalysisItem

Wraps analysis algo into data loading/saving and config.

make_params() → dict: Make global params for modeling.

abstract run(): See parent description.

abstract get_result(): See parent description.

class logml.analysis.items.modeling.EdaStep(*args, **kwargs)

Bases: logml.analysis.items.modeling.ModelingItemBase

EDA analysis

LABEL = 'run_eda'

PARAMS_CLS: alias of logml.analysis.items.modeling.ModelingConfigBase

run(): Run modeling step

get_result(): Modeling saves results directly, nothing to return here.

classmethod estimate_resources(res: logml.analysis.common.JobResourcesReqs, cfg: GlobalConfig = None, df: Optional[pandas.core.frame.DataFrame] = None, strata_shapes: Optional[Dict[str, tuple]] = None, item_params: Any = None) → None: See parent description

get_paths_to_release() → Optional[List[logml.analysis.base_item.ReleasePath]]: See parent description.

class logml.analysis.items.modeling.ModelingTransformerStepConfig

Bases: logml.analysis.items.modeling.ModelingConfigBase

Config for modelig dataset generation.

Show JSON schema

{
   "title": "ModelingTransformerStepConfig",
   "description": "Config for modelig dataset generation.",
   "type": "object",
   "properties": {
      "stratum_id": {
         "title": "Stratum Id",
         "type": "string"
      },
      "problem_id": {
         "title": "Problem Id",
         "type": "string"
      },
      "n_dataset": {
         "title": "N Dataset",
         "type": "integer"
      }
   },
   "required": [
      "stratum_id",
      "n_dataset"
   ]
}

Fields

n_dataset (int)

field n_dataset: int [Required]

class logml.analysis.items.modeling.ModelingTransformer(*args, **kwargs)

Bases: logml.analysis.items.modeling.ModelingItemBase

Wraps analysis algo into data loading/saving and config.

LABEL = 'modeling_data_transform'

PARAMS_CLS: alias of logml.analysis.items.modeling.ModelingTransformerStepConfig

run(): Run modeling step

get_result(): Modeling saves results directly, nothing to return here.

classmethod estimate_resources(res: logml.analysis.common.JobResourcesReqs, cfg: GlobalConfig = None, df: Optional[pandas.core.frame.DataFrame] = None, strata_shapes: Optional[Dict[str, tuple]] = None, item_params: Any = None) → None: See parent description

class logml.analysis.items.modeling.TrainModelStepConfig

Bases: logml.analysis.items.modeling.ModelingConfigBase

Config for model training.

Show JSON schema

{
   "title": "TrainModelStepConfig",
   "description": "Config for model training.",
   "type": "object",
   "properties": {
      "stratum_id": {
         "title": "Stratum Id",
         "type": "string"
      },
      "problem_id": {
         "title": "Problem Id",
         "type": "string"
      },
      "model_name": {
         "title": "Model Name",
         "type": "string"
      }
   },
   "required": [
      "stratum_id",
      "model_name"
   ]
}

Fields

model_name (str)

field model_name: str [Required]

class logml.analysis.items.modeling.TrainModelStep(*args, **kwargs)

Bases: logml.analysis.items.modeling.ModelingItemBase

Train Model Step.

LABEL = 'train_model'

PARAMS_CLS: alias of logml.analysis.items.modeling.TrainModelStepConfig

run(): Run modeling step

get_result(): Modeling saves results directly, nothing to return here.

classmethod estimate_resources(res: logml.analysis.common.JobResourcesReqs, cfg: GlobalConfig = None, df: Optional[pandas.core.frame.DataFrame] = None, strata_shapes: Optional[Dict[str, tuple]] = None, item_params: Any = None) → None: See parent description

class logml.analysis.items.modeling.SelectModelsStep(*args, **kwargs)

Bases: logml.analysis.items.modeling.ModelingItemBase

Models selection step.

LABEL = 'select_models'

PARAMS_CLS: alias of logml.analysis.items.modeling.ModelingConfigBase

run(): Run modeling step

get_result(): Modeling saves results directly, nothing to return here.

class logml.analysis.items.modeling.ExtractFeatureImportanceStepConfig

Bases: logml.analysis.items.modeling.ModelingConfigBase

Parameters for FI extraction.

Show JSON schema

{
   "title": "ExtractFeatureImportanceStepConfig",
   "description": "Parameters for FI extraction.",
   "type": "object",
   "properties": {
      "stratum_id": {
         "title": "Stratum Id",
         "type": "string"
      },
      "problem_id": {
         "title": "Problem Id",
         "type": "string"
      },
      "dataset_n": {
         "title": "Dataset N",
         "type": "integer"
      },
      "model_name": {
         "title": "Model Name",
         "type": "string"
      }
   },
   "required": [
      "stratum_id",
      "dataset_n",
      "model_name"
   ]
}

Fields

dataset_n (int)
model_name (str)

field dataset_n: int [Required]

field model_name: str [Required]

class logml.analysis.items.modeling.ExtractFeatureImportanceStep(*args, **kwargs)

Bases: logml.analysis.items.modeling.ModelingItemBase

FI extraction step.

LABEL = 'extract_fi'

PARAMS_CLS: alias of logml.analysis.items.modeling.ExtractFeatureImportanceStepConfig

run(): Run features selection step

get_result(): Modeling saves results directly, nothing to return here.

classmethod estimate_resources(res: logml.analysis.common.JobResourcesReqs, cfg: GlobalConfig = None, df: Optional[pandas.core.frame.DataFrame] = None, strata_shapes: Optional[Dict[str, tuple]] = None, item_params: Any = None) → None: See parent description.

class logml.analysis.items.modeling.ModelingAnalysisResult(selected_models: Optional[List[logml.model_search.common.ModelEvaluationData]] = None, fi_models: Optional[List[str]] = None, objective: Optional[logml.configuration.modeling.ModelingTaskSpec] = None, stratum_id: Optional[str] = None, problem_id: Optional[str] = None, num_features: int = - 1, top_features: Optional[List[str]] = None, analysis_name: Optional[str] = None, unique_name: Optional[str] = None, status: Optional[str] = None, fi_relative_loss: Optional[Dict[str, float]] = None)

Bases: logml.analysis.base_item.AnalysisResult

Modeling analysis result.

class logml.analysis.items.modeling.AggregateFeatureImportanceStep(*args, **kwargs)

Bases: logml.analysis.items.modeling.ModelingItemBase

FI Aggregation Step.

Note that this step is completely optional: in any case, all results are required. Here we only generate some sort of short summary table which is available even if reporting step fails fo some reason.

LABEL = 'combine_fi'

PARAMS_CLS: alias of logml.analysis.items.modeling.ModelingConfigBase

run(): See parent description.

generate_analysis_result(fi_runner, max_top_features_rank=30): Generates (high-level) modeling analysis result

get_result(): Modeling saves results directly, nothing to return here.

generate_analysis_metadata() → Optional[logml.analysis.base_item.AnalysisMetadata]

Creates metadata object.

Major use for metadata object is to point to high-level analysis results (for example, ultimate result of Survival Modeling, as opposed to sub-level analysis step, like model search). Those results are then gathered and rendered at the summary report page.

get_paths_to_release() → Optional[List[logml.analysis.base_item.ReleasePath]]: See parent description.

class logml.analysis.items.modeling.SurvivalAnalysisResult(stratum_id: Optional[str] = None, problem_id: Optional[str] = None, num_features: int = - 1, top_features: Optional[List[str]] = None, analysis_name: Optional[str] = None, unique_name: Optional[str] = None, status: Optional[str] = None)

Bases: logml.analysis.base_item.AnalysisResult

Survival step result

class logml.analysis.items.modeling.SurvivalAnalysisStep(*args, **kwargs)

Bases: logml.analysis.items.modeling.ModelingItemBase

Survival analysis step.

LABEL = 'survival_analysis'

PARAMS_CLS: alias of logml.analysis.items.modeling.ModelingConfigBase

run(): Run modeling step

generate_analysis_result(): Generate artifacts.

get_result(): Modeling saves results directly, nothing to return here.

generate_analysis_metadata() → Optional[logml.analysis.base_item.AnalysisMetadata]

Creates metadata object.

Major use for metadata object is to point to high-level analysis results (for example, ultimate result of Survival Modeling, as opposed to sub-level analysis step, like model search). Those results are then gathered and rendered at the summary report page.

class logml.analysis.items.modeling.GenerateReportStep(*args, **kwargs)

Bases: logml.analysis.items.modeling.ModelingItemBase

Generate Report Step.

LABEL = 'generate_report'

run(): Run modeling step

make_params() → dict: Make global params for modeling.

get_result(): See parent description.

classmethod estimate_resources(res: logml.analysis.common.JobResourcesReqs, cfg: GlobalConfig = None, df: Optional[pandas.core.frame.DataFrame] = None, strata_shapes: Optional[Dict[str, tuple]] = None, item_params: Any = None) → None: See parent description.

class logml.analysis.items.modeling.PackageReportStep(*args, **kwargs)

Bases: logml.analysis.items.modeling.ModelingItemBase

Package report step

LABEL = 'package_report'

run(): Run modeling step

make_params() → dict: Make global params for modeling.

get_result(): See parent description.

logml.analysis.items.modeling.stratum_high_resources_estimator(res: logml.analysis.common.JobResourcesReqs, strata_shapes: Optional[Dict[str, tuple]] = None, params: Optional[Any] = None, cpu_mul: float = 1.0, mem_mul: float = 2.0): Default estimation of high resources requirements based on dataset rows and columns number.