Source code for pm4py.algo.discovery.dfg.adapters.pandas.freq_triples

from pm4py.util import constants, pandas_utils




def get_freq_triples(
    df,
    activity_key="concept:name",
    case_id_glue="case:concept:name",
    timestamp_key="time:timestamp",
    sort_caseid_required=True,
    sort_timestamp_along_case_id=True,
):
    """
    Gets the frequency triples out of a dataframe

    Parameters
    ------------
    df
        Dataframe
    activity_key
        Activity key
    case_id_glue
        Case ID glue
    timestamp_key
        Timestamp key
    sort_caseid_required
        Determine if sort by case ID is required (default: True)
    sort_timestamp_along_case_id
        Determine if sort by timestamp is required (default: True)

    Returns
    -------------
    freq_triples
        Frequency triples from the dataframe
    """
    import pandas as pd

    if sort_caseid_required:
        if sort_timestamp_along_case_id:
            df = df.sort_values([case_id_glue, timestamp_key])
        else:
            df = df.sort_values(case_id_glue)
    df_reduced = df[[case_id_glue, activity_key]]
    # shift the dataframe by 1
    df_reduced_1 = df_reduced.shift(-1)
    # shift the dataframe by 2
    df_reduced_2 = df_reduced.shift(-2)
    # change column names to shifted dataframe
    df_reduced_1.columns = [str(col) + "_2" for col in df_reduced_1.columns]
    df_reduced_2.columns = [str(col) + "_3" for col in df_reduced_2.columns]

    df_successive_rows = pandas_utils.concat(
        [df_reduced, df_reduced_1, df_reduced_2], axis=1
    )
    df_successive_rows = df_successive_rows[
        df_successive_rows[case_id_glue]
        == df_successive_rows[case_id_glue + "_2"]
    ]
    df_successive_rows = df_successive_rows[
        df_successive_rows[case_id_glue]
        == df_successive_rows[case_id_glue + "_3"]
    ]
    all_columns = set(df_successive_rows.columns)
    all_columns = list(
        all_columns
        - set([activity_key, activity_key + "_2", activity_key + "_3"])
    )
    directly_follows_grouping = df_successive_rows.groupby(
        [activity_key, activity_key + "_2", activity_key + "_3"]
    )
    if all_columns:
        directly_follows_grouping = directly_follows_grouping[all_columns[0]]
    freq_triples = directly_follows_grouping.size().to_dict()
    return freq_triples