Add F1 score computation from precision recall curve

results/roc.py

@@ -7,12 +7,15 @@ import subprocess
 import matplotlib.pyplot as plt
 import numpy as np
 import pandas as pd
+from warnings import simplefilter
 
-from sklearn.metrics import roc_auc_score, roc_curve, f1_score
+from sklearn.metrics import roc_auc_score, precision_recall_curve
 from sklearn.preprocessing import Normalizer
 
 from .base import BaseResults
 
+simplefilter("ignore", category=RuntimeWarning)
+
 class ROCResults(BaseResults):
     """
     This class will compute the ROC metric.
@@ -107,19 +110,11 @@ class ROCResults(BaseResults):
 
         # Once the correct anomaly scores have been computed, we can compute the metrics
         roc = roc_auc_score(labels, result)
-        fpr, tpr, thresh = roc_curve(labels, result)
-
-        closest_dist = np.inf
-        closest_thresh = np.inf
-        best_couple = (0,0)
-        for f, t, th in zip(fpr, tpr, thresh):
-            dist = np.sqrt((f-0)**2+(t-1)**2)
-            if dist < closest_dist:
-                closest_dist = dist
-                closest_thresh = th
-                best_couple = (f, t)
-
-        binary_labels = (result > closest_thresh).astype(int)
-        f1 = f1_score(labels, binary_labels)
-
-        return roc, f1
+        prec, rec, _ = precision_recall_curve(labels, result)
+
+        fscore = (2 * prec * rec) / (prec + rec)
+        fscore = np.nan_to_num(fscore)
+
+        idx = np.argmax(fscore)
+
+        return roc, fscore[idx]