Adversarial classification

import pandas as pd
from upsetplot import UpSet, from_memberships
import matplotlib.pyplot as plt

# Define archetypal memberships.
archetypes_definitions = [
    {"groups": ["Israeli Citizen", "Jew", "Likud"], "count": 15},
    {"groups": ["Israeli Citizen", "Jew", "WB Resident"], "count": 7},
    {"groups": ["Israeli Citizen", "Jew"], "count": 25},
    {"groups": ["Israeli Citizen", "Arab", "Muslim", "Sunni"], "count": 12},
    {"groups": ["Israeli Citizen", "Druze", "Arab"], "count": 3},
    {"groups": ["Israeli Citizen", "Druze"], "count": 2},
    {"groups": ["Israeli Citizen", "Arab", "Bedouin", "Muslim", "Sunni"], "count": 4},
    {"groups": ["Israeli Citizen", "Arab", "Christian"], "count": 3},
    {
        "groups": [
            "Palestinian",
            "WB Resident",
            "Arab",
            "Muslim",
            "Sunni",
            "PA Affiliate",
        ],
        "count": 18,
    },
    {
        "groups": ["Palestinian", "Gaza Resident", "Arab", "Muslim", "Sunni", "Hamas"],
        "count": 12,
    },
    {"groups": ["Palestinian", "WB Resident", "Arab", "Christian"], "count": 2},
    {
        "groups": ["Palestinian", "Gaza Resident", "Arab", "Muslim", "Sunni"],
        "count": 25,
    },
    {"groups": ["Palestinian", "WB Resident", "Arab", "Muslim", "Sunni"], "count": 20},
    {"groups": ["Hezbollah", "Muslim", "Shiite", "Arab"], "count": 8},
    {"groups": ["Houthi", "Muslim", "Shiite", "Arab"], "count": 6},
]

memberships_data = []
for archetype in archetypes_definitions:
    current_groups = list(archetype["groups"])
    if "WB Resident" in current_groups or "Gaza Resident" in current_groups:
        if "OT Resident" not in current_groups:
            current_groups.append("OT Resident")
    if "Muslim" in current_groups and not (
        "Sunni" in current_groups or "Shiite" in current_groups
    ):
        if not any(
            group_name in current_groups for group_name in ["Hezbollah", "Houthi"]
        ):
            current_groups.append("Sunni")
    for _ in range(archetype["count"]):
        memberships_data.append(current_groups)

# Create the data structure for UpSet plot
# from_memberships with this input type will return a pandas Series
# where the index is a MultiIndex of booleans (sets), and values are 1s.
# The index will NOT be unique if archetypes are repeated.
upset_data_series = from_memberships(memberships_data)

## Debugging
# print(f"Type of data from from_memberships: {type(upset_data_series)}")
# if isinstance(upset_data_series, pd.Series) and isinstance(
#     upset_data_series.index, pd.MultiIndex
# ):
#     print(f"Series head (index may not be unique yet):\n{upset_data_series.head()}")
#     print(f"Index names: {upset_data_series.index.names}")
#     print(
#         f"Total archetypal individuals represented (length of series): {len(upset_data_series)}"
#     )
#     print(f"Is index unique before UpSet: {upset_data_series.index.is_unique}")

# else:
#     print(
#         f"Warning: Data from from_memberships is not the expected Series with MultiIndex. Type: {type(upset_data_series)}"
#     )
#     if isinstance(upset_data_series, pd.Series):
#         print(f"Series index type: {type(upset_data_series.index)}")

final_data_for_plot = upset_data_series

try:
    # Create the plot
    # Set subset_size='sum' to aggregate the 'ones' from from_memberships.
    upset_plot_object = UpSet(
        final_data_for_plot,
        subset_size="sum",
        min_subset_size=1,
        show_counts=True,
        sort_by="cardinality",  # Sorts intersection bars
        sort_categories_by="cardinality",  # Sorts the sets on the left
    )

    # Styling subsets: 'present' should use names that are in upset_data_series.index.names
    if "OT Resident" in upset_data_series.index.names:
        upset_plot_object.style_subsets(
            present=["OT Resident"],
            facecolor="lightblue",
            label="In Occupied Territories",
        )
    if "Hamas" in upset_data_series.index.names:
        upset_plot_object.style_subsets(
            present=["Hamas"],
            edgecolor="red",
            hatch="xx",
            linewidth=2,
            label="Hamas Affiliation",
        )
    if "Likud" in upset_data_series.index.names:
        upset_plot_object.style_subsets(
            present=["Likud"],
            edgecolor="blue",
            hatch="//",
            linewidth=2,
            label="Likud Affiliation",
        )

    fig = plt.figure(
        # figsize=(18, 10)
    )
    upset_plot_object.plot(fig=fig)

    # plt.suptitle(
    #     "Conceptual Group Memberships In and Around Israel-Palestine",
    #     fontsize=16,
    # )
    plt.figtext(
        0.5,
        0.01,
        "NOTE: Intersection sizes are based on defined archetypes, not actual population data. This is a conceptual model.",
        ha="center",
        fontsize=8,
        style="italic",
    )
    plt.tight_layout(rect=[0, 0.03, 1, 0.95])
    # plt.savefig("israel_palestine_upset_plot.png", dpi=300)
    # plt.show()

    # print("UpSet plot 'israel_palestine_upset_plot.png' generated.")
    # Access the aggregated intersections from the UpSet object for the count
    # aggregated_intersections = upset_plot_object.intersections
    # print(
    #     f"Number of unique intersections shown: {len(aggregated_intersections[aggregated_intersections >= 1])}"
    # )
    # print(f"Total sum of intersection sizes: {aggregated_intersections.sum()}")


except Exception as e:
    print(f"An error occurred during plot generation: {e}")
    print(
        f"Data passed to UpSet (head):\n {final_data_for_plot.head() if isinstance(final_data_for_plot, pd.Series) else 'Data not a Series'}"
    )
    if isinstance(final_data_for_plot, pd.Series) and isinstance(
        final_data_for_plot.index, pd.MultiIndex
    ):
        print(f"Index names of data: {final_data_for_plot.index.names}")