Modalities package

Submodules

sax.core.causal_process_discovery.modalities.base_anchor module

class sax.core.causal_process_discovery.modalities.base_anchor.BaseAnchor

Bases: object

Parent class for all causal discovery modalities. Please extend this class to add more modalities. Different modalities are described here <>

abstract apply(dataObject: RawEventData, variant: Algorithm | None = Algorithm.POSITIVE_LINGAM, prior_knowledge: bool | None = False, threshold: float | None = 0.5) → CausalResultInfo

Apply the chosen algorithm variant with the current modality on the process data object.

Parameters:

dataObject (RawEventData) -- event log
variant (Optional[Algorithm], optional) -- the chosen algorithm variant to apply, defaults to LINGAM

Returns:

causal discovery result

Return type:

CausalResultInfo

sax.core.causal_process_discovery.modalities.chain_anchor module

class sax.core.causal_process_discovery.modalities.chain_anchor.ChainAnchorTransformer

Bases: BaseAnchor

Causal chain modality transformer - transforms process data object from event log format to a tabular format which represents each activity as a column in the data frame, where the value of the column is the accumulated duration of this activity from trace start time. The chain modality is described here <>

apply(dataObject: RawEventData, variant: Algorithm | None = Algorithm.POSITIVE_LINGAM, prior_knowledge: bool | None = False, threshold: float | None = 0.5) → CausalResultInfo

Transform raw event log in tabular form (the provided event log) to row format, so that each trace is represented by a row and each activity column value holds the accimulated duration from trace start time. After the transformation, applies the chosen causal discovery algorithm variant to discover causal execution dependencies among chained durations.

Parameters:

dataObject (RawEventData) -- event log
variant (Optional[Algorithm], optional) -- algorithm variant to apply, defaults to LINGAM
prior_knowledge (Optional[bool], optional) -- whether to apply prior knowledge, defaults to False

Raises:

TypeError -- _description_

Returns:

_description_

Return type:

CausalResultInfo

sax.core.causal_process_discovery.modalities.parent_anchor module

class sax.core.causal_process_discovery.modalities.parent_anchor.ParentAnchorTransformer

Bases: BaseAnchor

Parent Causal Modality - methods to apply causal discovery on event log with PARENT modality

apply(dataObject: RawEventData, variant: Algorithm | None = Algorithm.POSITIVE_LINGAM, prior_knowledge: bool | None = False, depth: int = 1) → CausalResultInfo

Transpose the original event log so that each column is activity name and the value is the timestamp of the activity. Calculate all possible pairs of activities within this process and all its variants which happen one after the other. For each part, across all variants holding this pair, compute the time diff with the parent activity of the pair. Apply causal discovery algorithm of the subframe defined by the parent activity and the pair of activities and the time diff between them. This will be a single coefficient in the adjacenty matrix of the whole process. At the end create global adjacency matrix out of all coefficients.

Parameters

dataObjectRawEventData: event log data object
variantOptional[Algorithm], optional: algorithm variant to apply, defaults to LINGAM
prior_knowledgeOptional[bool], optional: whether to apply prior knowledge, defaults to False
depthint, optional: depth of the distance between activities to consider as pair, defaults to 1

Raises

TypeError: The method can be applied only to an object of type RawEventData!

Returns

CausalResultInfo: causal discovery result