Loading Data

import pandas as pd

# Load and drop unwanted index column if present
def load_top100(path):
    s = pd.read_csv(path, header=None)[0]
    s = s[s != "0"]      # ⬅ remove bogus index-row
    return s.tolist()

lasso_auc_top100 = load_top100("../results/lasso_auc_top100.csv")
lasso_recall_top100 = load_top100("../results/lasso_recall_top100.csv")

en_auc_top100 = load_top100("../results/elasticnet_auc_top100.csv")
en_recall_top100 = load_top100("../results/elasticnet_recall_top100.csv")

rfe_auc_top100 = load_top100("../results/rfe_auc_top100.csv")
rfe_recall_top100 = load_top100("../results/rfe_recall_top100.csv")

rfecv_auc_top100 = load_top100("../results/rfecv_auc_top100.csv")
rfecv_recall_top100 = load_top100("../results/rfecv_recall_top100.csv")

Building the Master Method Dictionary

methods = {
    "Lasso AUC": lasso_auc_top100,
    "Lasso Recall": lasso_recall_top100,

    "EN AUC": en_auc_top100,
    "EN Recall": en_recall_top100,

    "RFE AUC": rfe_auc_top100,
    "RFE Recall": rfe_recall_top100,

    "RCV AUC": rfecv_auc_top100,
    "RCV Recall": rfecv_recall_top100,
}

Building the Unified Feature Set

all_feats = sorted(set().union(*methods.values()))

Constructing the Agreement Table

table = pd.DataFrame({"Feature": all_feats})

for m in methods:
    table[m] = table["Feature"].apply(lambda f: 1 if f in methods[m] else 0)

Sum of how many methods selected each feature

table["SUM"] = table[list(methods.keys())].sum(axis=1)

Sort by agreement strength

table = table.sort_values("SUM", ascending=False)

Exporting the Table

table.to_csv("../results/feature_agreement_table.csv", index=False)

print("Agreement table saved → ../results/feature_agreement_table.csv")

Agreement table saved → ../results/feature_agreement_table.csv