add new file for roc compering models, add cells to roc_curve

Author: ksopyla

Diff for: metrics/roc_curve.py

@@ -1,22 +1,29 @@
+"""Example of computing ROC curve
+"""
+#%%
 import numpy as np
-from sklearn import metrics
 import matplotlib.pyplot as plt
+import sklearn.metrics as skm
 
+#y_true = np.array([0, 0, 0, 1, 0, 1, 1, 1])
 y_true = np.array([0, 0, 0, 0, 0, 1, 1, 1])
 y_score = np.array([0.1, 0.4, 0.35, 0.7, 0.2, 0.3, 0.6, 0.8])
 
-fpr, tpr, tresholds = metrics.roc_curve(y_true, y_scores, pos_label=1)
+#%%
+fpr, tpr, tresholds = skm.roc_curve(y_true, y_score, pos_label=1)
 
-print(f'\nfpr={fpr}\ntpr={tpr}\ntre={thresholds}')
+print(f'\nfpr={fpr}\ntpr={tpr}\ntre={tresholds}')
 
 
 for tr, tp, fp in zip(tresholds, tpr, fpr):
-    y_pred = (y_scores>=tr).astype(int)
-    cm = metrics.confusion_matrix(y_true, y_pred)
+    y_pred = (y_score>=tr).astype(int)
+    cm = skm.confusion_matrix(y_true, y_pred)
     print(cm)
-    print(metrics.classification_report(y_true, y_pred))
+    print(skm.classification_report(y_true, y_pred))
     print(f'Treshold={tr},tpr={tp}, fpr={fp} predicions={y_pred}')
 
+
+#%%
 plt.plot(fpr, tpr, lw=1)
 plt.scatter(fpr,tpr)
 
@@ -26,4 +33,6 @@
 plt.ylabel('True Positive Rate')
 plt.title('Receiver operating characteristic example')
 #plt.legend(loc="lower right")
-plt.show()
+plt.show()
+
+# %%

Diff for: metrics/roc_curve_model_comparision.py

@@ -0,0 +1,64 @@
+#%%
+import matplotlib.pyplot as plt
+import numpy as np
+import sklearn.metrics as skm
+
+
+
+y_true = np.array([0, 0, 0, 0, 1, 1, 1, 1, 1, 1])
+
+# model1 curve will dominate the model0
+# output from model0
+y_score0 = np.array([0.75, 0.5, 0.3, 0.35, 0.45, 0.7, 0.3, 0.33, 0.5, 0.8]) 
+# output from model1
+y_score1 = np.array([0.6, 0.3, 0.3, 0.55, 0.65, 0.4, 0.55, 0.33, 0.75, 0.3])
+
+# model 1 is better
+# # output from model0
+# y_score0 = np.array([0.75, 0.5, 0.3, 0.35, 0.45, 0.7, 0.3, 0.33, 0.5, 0.8])
+# # output from model1
+# y_score1 = np.array([0.7, 0.3, 0.3, 0.55, 0.75, 0.4, 0.5, 0.33, 0.72, 0.3])
+
+# looking only at curves it is not so obvious, which one is better
+# output from model0
+# y_score0 = np.array([0.7, 0.45, 0.3, 0.35, 0.45, 0.7, 0.3, 0.33, 0.55, 0.8])
+# # output from model1
+# y_score1 = np.array([0.6, 0.3, 0.3, 0.55, 0.65, 0.4, 0.5, 0.33, 0.75, 0.3])
+
+
+# %
+
+
+# first model
+fpr0, tpr0, tresholds0 = skm.roc_curve(y_true, y_score0)
+
+# second model
+
+fpr1, tpr1, tresholds1 = skm.roc_curve(y_true, y_score1)
+
+
+auc_roc0 = skm.roc_auc_score(y_true, y_score0)
+auc0 = skm.auc(fpr0,tpr0)
+print(f"Model 0 auc roc={auc_roc0} area under curve={auc0}")
+
+auc_roc1 = skm.roc_auc_score(y_true, y_score1)
+auc1 = skm.auc(fpr1,tpr1)
+print(f"Model 0 auc roc={auc_roc1} area under curve={auc1}")
+
+
+# % plot curve
+plt.plot(fpr0,tpr0, 'ro')
+plt.plot(fpr0,tpr0, 'r', label='model 0')
+
+plt.plot(fpr1,tpr1, 'bo')
+plt.plot(fpr1,tpr1, 'b', label='model 1')
+
+plt.xlabel('FPR')
+plt.ylabel('TPR')
+plt.ylim([0.0, 1.05])
+plt.xlim([0.0, 1.0])
+plt.title('ROC curve for 2 ml models')
+plt.legend()
+plt.show()
+
+# %%