API

Training

This module is responsible for the training phase of ECS Event Abstraction.

ecsea.training.train_model(low_level_log: DataFrame, high_level_log: DataFrame, parameter: Dict[str, Any] | None = None, show_progress=True) → Model

Train a model by “comparing” a low-level and a high-level event log. The low-level and high-level traces should contain the same traces with the same IDs as “concept:name” attribute.

Parameters:

  • low_level_log – the low-level event log

  • high_level_log – the high-level event log

  • parameter – additional training parameter

  • show_progress – show a progress bar

Returns:

the Model

ecsea.training.validate_logs(low_level_log: List[List[Dict[Hashable, Any]]], high_level_log: List[List[Dict[Hashable, Any]]])

Tests if the low-level and the high-level log are suitable for this ecs event abstraction approach. The function raises a ValueError, if the high-level log contains a trace, which is not contained in the low-level log.

Parameters:

  • low_level_log – the low level log

  • high_level_log – the high level log

Returns:

nothing

Parameter

This module provides functions for the ecsea parameter.

exception ecsea.parameter.EcsEaParameterError(parameter_key, message)
class ecsea.parameter.ParameterKey(*values)
ecsea.parameter.get_default_parameter() → Dict

Returns a dict of default parameters for the ECS-EA training algorithm.

ecsea.parameter.validate_parameter(parameter: Dict)

Validate the parameters and throws a Value Error, if parameters does not fit the constraints.

Model

class ecsea.model.Model

Defines an ECSEA model, that was trained with a low-level and a high-level log. Can be applied with a new low-level log to create a high-level log.

export(path: str)

Export the model to the given file path.

Parameters:

path – The path for the model file

static load(path: str) → Model

Load the model from the given file path.

Parameters:

path – The path for the model file

pretty_string(show_activities=True, show_accuracy_information=False, show_parameter=True, show_mapping_information=True, show_mapping_statistics=True) → str

Create a string representation of the model the information stored in this model.

Parameters:

  • show_activities – Show the low-level and high-level activities stored in this model

  • show_accuracy_information – Show the accuracy information from the training phase

  • show_parameter – Show the parameter for the training & application phase

  • show_mapping_information – Show the mapping information stored in the model

  • show_mapping_statistics – Show the statistics for the mapping information

Application

This module is responsible for the application phase of ECS Event Abstraction.

class ecsea.application.ApplicationResult(abstracted_log: pandas.core.frame.DataFrame, compression: float, runtime_ms: float)
ecsea.application.apply(low_level_log: DataFrame, model: Model, show_progress=False) → ApplicationResult

Converts a low-level event log into a high-level event log using a trained model. This model was created by training it with a low-level event log and the corresponding high-level event log.

Parameters:

  • low_level_log – the low-level event log

  • model – the trained model

  • show_progress – show a progress bar

Returns:

the abstracted event log

ecsea.application.apply_trace(low_level_trace: List[Dict[Hashable, Any]], model: Model) → List[Dict[Hashable, Any]]

Convert a single trace into a high-level trace

Parameters:

  • low_level_trace – the low-level trace

  • model – the Model

Returns:

a new high-level trace based on the converted low-level trace

ecsea.application.apply_trace_old(low_level_trace: List[Dict[Hashable, Any]], model: Model) → List[Dict[Hashable, Any]]

Convert a single trace into a high-level trace

Parameters:

  • low_level_trace – the low-level trace

  • model – the Model

Returns:

a new high-level trace based on the converted low-level trace

Optimize

This module is responsible for optimizing the training phase of ECS Event Abstraction.

ecsea.optimize.optimize(low_level_log, high_level_log, parameter: Dict[str, Any], accuracy_attribute='train_accuracy_dl', n_trials=100, timeout=None, *, optimize_time_span=True, optimize_timestamp_merge=True, optimize_mapping_threshold=True, min_time_span=500, max_time_span=100000, min_mapping_threshold=0.0, max_mapping_threshold=3.0)

Optimize the model by performing a hyperparamter optimization.

Parameters:

  • low_level_log – The low-level log

  • high_level_log – The high-level log

  • parameter – the default parameter for the optimization

  • accuracy_attribute – the model attribute for the accuracy {see POSSIBLE_ACCURACY_ATTRIBUTES}

  • n_trials – the number of trials

  • timeout – the timeout, if None -> No timeout

  • optimize_time_span – Optimize the time span; default: True

  • optimize_timestamp_merge – Optimize the timestamp_merge; default: True

  • optimize_mapping_threshold – Optimize the mapping_threshold; default: True

  • min_time_span – The upper bound for the time_span parameter

  • max_time_span – The lower bound for the time_span parameter

  • min_mapping_threshold – The lower bound for the mapping_threshold (should be >=0

  • max_mapping_threshold – The upper bound for the mapping_threshold

Returns:

The best params for the best model