From 740a5e559b39fd3738b603c209fc4b1a77a5425b Mon Sep 17 00:00:00 2001
From: "tobias.rechberger" <tobias.rechberger@crayon.com>
Date: Tue, 21 Jul 2020 12:21:39 +0200
Subject: [PATCH] Added individual scores per attributes of an ovservation and
 adapted example acc.

---
 coupled_biased_random_walks/detect.py | 29 ++++++++++++++++++++++++++-
 example.py                            | 11 ++++++++--
 2 files changed, 37 insertions(+), 3 deletions(-)

diff --git a/coupled_biased_random_walks/detect.py b/coupled_biased_random_walks/detect.py
index e4c5f3c..cca76a8 100644
--- a/coupled_biased_random_walks/detect.py
+++ b/coupled_biased_random_walks/detect.py
@@ -91,13 +91,40 @@ def score(self, observation_iterable):
 
     def _score(self, observation):
         """
-        Computes the weighted anomaly score for an observation
+        Computes the weighted anomaly score (object_score in the paper) for an observation
         :param observation: dict of the form {feature_name: feature_value, ...}
         """
         return sum(self._get_feature_relevance(item) * \
                    self._stationary_prob.get(item, self._unknown_feature_score)
                     for item in iteritems(observation))
 
+    def value_scores(self, observation_iterable):
+        """
+        Compute an anomaly sub-score for each value of the observation in observation_iterable
+        :param observation_iterable: iterable of dict observations with each dict
+        of the form {feature_name: feature_value, ...}
+        Return dict with sub score of each value of the observation/object of the form:
+        {feature_name: weighted score of value of feature, ...}
+        (sub-scores sum up to score(self, observation_iterable))
+        """
+        if not (self._feature_relevance and self._stationary_prob):
+            raise CBRWScoreError()
+        if isinstance(observation_iterable, dict):
+            observation_iterable = [observation_iterable]
+        return np.array([self._value_scores(obs) for obs in observation_iterable])
+
+    def _value_scores(self, observation):
+        """
+        Computes the weighted value scores for each feature value of an observation
+        """
+        score_keys = []
+        value_scores = []
+        for item in iteritems(observation):
+            score_keys.append(item[0])
+            value_scores.append(self._get_feature_relevance(item) * \
+                               self._stationary_prob.get(item, self._unknown_feature_score))
+        return dict(zip(score_keys, value_scores))
+
     def _get_feature_relevance(self, feature_tuple):
         """
         Getter for the relevance (weight) of a feature (category)
diff --git a/example.py b/example.py
index f151359..a88672b 100644
--- a/example.py
+++ b/example.py
@@ -2,7 +2,8 @@
 
 from six.moves import zip
 
-from coupled_biased_random_walks import CBRW
+#from coupled_biased_random_walks import CBRW
+from coupled_biased_random_walks.detect import CBRW
 from data.loading import load_from_csv
 
 
@@ -22,8 +23,14 @@
 
     # fit and score data
     detector.fit()
+    print('Feature weights: ', detector.feature_weights)
+    print()
     scores = detector.score(data)
-
     # print scores and observations
     for score, datum in zip(scores, data):
         print('Score: {} | Data: {}'.format(round(score, 4), datum))
+
+    print()
+    print('Scores: ', scores)
+    print()
+    print('Value scores per attribute: ', detector.value_scores(data))
\ No newline at end of file