Paper: Computational Resource Optimisation in Feature Selection under Class Imbalance Conditions

papers/amadi_udu/codes/box_plots.py

@@ -0,0 +1,63 @@
+#!/usr/bin/env python
+# coding: utf-8
+# %%
+
+# %%
+
+
+import pandas as pd
+import ast
+import matplotlib.pyplot as plt
+import numpy as np
+
+
+# %%
+
+
+#choose dataset and model from list
+datasets = {0: 'census_income', 1: 'bank_marketing', 2: 'statlog_shuttle', 3: 'diabetes', 4:'gsvs', 5: 'cover_type'}
+models= {0: 'lgbm', 1: 'rf', 2: 'svm'}
+
+model = models[0]
+dataset = datasets[0]
+
+
+# %%
+
+
+df = pd.read_csv(f'results/{model}/{model}_{dataset}_combined_importances.csv')
+df['AUC'] = df['AUC'].apply(lambda x: ast.literal_eval(x))
+
+number = 0
+y_ticks = np.round(np.arange(0.1, 1.1, 0.1), 2).tolist()
+xmin, xmax = round(min(min(df['AUC'])),2)-0.05, round(max(max(df['AUC'])),2)+0.1
+for _ in range(int(df.shape[0]/10)):
+    f_no = df['Feature_ID'].iloc[0+number]
+    plt.figure()
+    df['AUC'].iloc[0+number:10+number].apply(lambda x: pd.Series(x)).T.boxplot(figsize=(7, 5), vert=False,
+                                                            boxprops=dict(color='blue'),
+                                                            medianprops=dict(color='green'),
+                                                            flierprops=dict(marker='x', markersize=6, markeredgecolor='red'),
+                                                            whiskerprops=dict(linestyle='--', dashes = (10, 5)))
+
+    plt.rc('font', family='Times New Roman')
+    plt.xlabel('Decrease in AUC',fontsize=14)
+    plt.ylabel('Sample Fractions',fontsize=14)
+    plt.yticks(ticks=np.arange(1, 11), labels=y_ticks)
+
+    plt.xlim([xmin, xmax])
+    plt.axvline(x=0, color='black', linestyle='-.')
+    plt.tick_params(axis='both', which='major', labelsize=14)
+    plt.grid(visible=True)
+    plt.tight_layout()
+    plt.savefig(f'results/{model}/{model}_{dataset}_feature_{f_no}.png', dpi=300, bbox_inches='tight')
+    plt.clf()
+    plt.close()
+    number += 10
+
+
+# %%
+
+
+
+