Source code for pm4py.algo.organizational_mining.sna.util

from typing import List, Dict
from enum import Enum
from pm4py.util import exec_utils, nx_utils
from pm4py.objects.org.sna.obj import SNA
import numpy as np




class Parameters(Enum):
    WEIGHT_THRESHOLD = "weight_threshold"





def sna_result_to_nx_graph(sna: SNA, parameters=None):
    """
    Transforms the results of SNA to a NetworkX Graph / DiGraph object
    (depending on the type of analysis).

    Parameters
    ------------------
    sna
        Result of a SNA operation
    parameters
        Parameters of the algorithm, including:
        - Parameters.WEIGHT_THRESHOLD => the weight threshold (used to filter out edges)

    Returns
    -----------------
    nx_graph
        NetworkX Graph / DiGraph
    """
    if parameters is None:
        parameters = {}

    weight_threshold = exec_utils.get_param_value(
        Parameters.WEIGHT_THRESHOLD, parameters, 0.0
    )
    directed = sna.is_directed

    if directed:
        graph = nx_utils.DiGraph()
    else:
        graph = nx_utils.Graph()

    graph.add_edges_from(
        {c for c, w in sna.connections.items() if w >= weight_threshold}
    )

    return graph





def cluster_affinity_propagation(
    sna: SNA, parameters=None
) -> Dict[str, List[str]]:
    """
    Performs a clustering using the affinity propagation algorithm provided by Scikit Learn

    Parameters
    --------------
    sna
        Result of a SNA operation
    parameters
        Parameters of the algorithm

    Returns
    --------------
    clustering
        Dictionary that contains, for each cluster that has been identified,
        the list of resources of the cluster
    """
    from pm4py.util import ml_utils

    if parameters is None:
        parameters = {}

    originators = list(
        set(x[0] for x, y in sna.connections.items()).union(
            set(x[1] for x, y in sna.connections.items())
        )
    )
    matrix = np.zeros((len(originators), len(originators)))
    for c, w in sna.connections.items():
        matrix[originators.index(c[0]), originators.index(c[1])] = w

    affinity_propagation = ml_utils.AffinityPropagation(**parameters)
    affinity_propagation.fit(matrix)

    clusters = affinity_propagation.predict(matrix)
    ret = {}
    for i in range(len(clusters)):
        res = originators[i]
        cluster = str(clusters[i])
        if cluster not in ret:
            ret[cluster] = []
        ret[cluster].append(res)

    return ret