Causal Discovery Algorithms package

This package contains the relevant causal discovery algorithms that can be applied on process data to discover causal relationships between activities within the process. The current algorithm implementation include DirectLINGAM algorithm. Additional causal discovery algorithms can be added to the package as long as they extend the sax.causal_process_discovery.algorithms.base_causal_alg.BaseCausalAlgorithm class.

Submodules

sax.core.causal_process_discovery.algorithms.base_causal_alg module

class sax.core.causal_process_discovery.algorithms.base_causal_alg.BaseCausalAlgorithm(data: DataFrame, prior_knowledge: PriorKnowledge | None = None)

Bases: object

Parent class for all causal discovery algorithms implementations. Please extend this class to add more algorithms.

abstract run() → ResultInfo

Run the causal discovery algorithm and return the result

Returns

return:

A object holding the causal discovery result

rtype:

ResultInfo

Raises

raises NotImplementedError:

abstract sanity_check()

Sanity check of the data object for causal discovery

Returns

return:

Boolean indicator whether the data object is coherent

rtype:

bool

Raises

raises NotImplementedError:

exception sax.core.causal_process_discovery.algorithms.base_causal_alg.CausalDataException(message)

Bases: Exception

Exception indicating that the data object containing process data is not suitable for causal discovery in its present form.

Parameters:

Exception -- exception indicating the reason why the process data object is not suitable for causal discovery

class sax.core.causal_process_discovery.algorithms.base_causal_alg.CausalResultInfo(adjacencyMatrix: List[List[int]], columns: List[str])

Bases: ResultInfo

Class containing causal discovery results. Basically this is a wrapper around causal adjacency matrix.

getAdjacencyMatrix() → List[List[int]]

Return causal adjacency matrix contained within this object

Returns:

causal adjacency matrix

Return type:

List[List[int]]

getColumns() → List[str]

Return a list with the names of the columns appearing in the adjacency matrix

Returns:

List of column names

Return type:

List[str]

getDiscoveryResult() → str

Return a NL description of the causal relationships

Returns:

Textual description of causal relationships

Return type:

str

sax.core.causal_process_discovery.algorithms.lingam module

class sax.core.causal_process_discovery.algorithms.lingam.LingamImpl(data: DataFrame, prior_knowledge: PriorKnowledge | None = None)

Bases: BaseCausalAlgorithm

DirectLINGAM causal discovery algorithm wrapper for process execution causal discovery. Make use of DirectLINGAM algorithm

Parameters:

BaseCausalAlgorithm (BaseCausalAlgorithm) -- base type

run() → CausalResultInfo

Runs DirectLINGAM on the data the algorithm instance was initiated with, and returns result object containing LINGAM adjaceny matrix.

Returns:

CausalResultInfo

Return type:

CausalResultInfo

sanity_check()

Sanity check for LINGAM, validates that the data on which it is run does not contain NaN values.

Raises:

CausalDataException -- throws exception in case data contains NaN values.

Module contents