Skip to content

noveltydetectionclassifier

NoveltyDetectionClassifier

Bases: MiniBatchClassifier

Source code in streamndr/model/noveltydetectionclassifier.py 
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
class NoveltyDetectionClassifier(base.MiniBatchClassifier):

    def __init__(self,
                 verbose=0,
                 random_state=None):
        super().__init__()
        self.nb_class_unknown = dict()
        self.class_sample_counter = dict()
        self.sample_counter = 0
        self.verbose = verbose
        self.random_state = random_state
        self.before_offline_phase = True

    def get_unknown_rate(self):
        """Returns the unknown rate, represents the percentage of unknown samples on the total number of samples classified in the online phase.

        Returns
        -------
        float
            Unknown rate
        """
        if self.sample_counter == 0:
            return 0.0
        return len(self.short_mem) / self.sample_counter

    def get_class_unknown_rate(self):
        """Returns the unknown rate per class. Represents the percentage of unknown samples on the total number of samples of that class seen during the stream.

        Returns
        -------
        dict
            Dictionary containing the unknown rate of each class
        """
        rates = {}
        for key, val in self.nb_class_unknown.items():
            total = self.class_sample_counter.get(key, 0)
            rates[key] = (val / total) if total > 0 else 0.0
        return rates

    def predict_proba_one(self,X):
        #Function used by river algorithms to get the probability of the prediction. It is not applicable to this algorithm since it only predicts labels. 
        #It is only added as to follow River's API.
        pass

    def predict_proba_many(self, X):
        #Function used by river algorithms to get the probability of the predictions. It is not applicable to this algorithm since it only predicts labels. 
        #It is only added as to follow River's API.
        pass

    def _remove_sample_from_short_mem(self, index):
        instance = self.short_mem.get_instance(index)
        if instance is None:
            return
        y_true = instance.y_true
        if y_true is not None:
            self.nb_class_unknown[y_true] -= 1
        self.short_mem.remove(index)

get_class_unknown_rate()

Returns the unknown rate per class. Represents the percentage of unknown samples on the total number of samples of that class seen during the stream.

Returns:

Type Description
dict

Dictionary containing the unknown rate of each class
Source code in streamndr/model/noveltydetectionclassifier.py 
30
31
32
33
34
35
36
37
38
39
40
41
42
def get_class_unknown_rate(self):
    """Returns the unknown rate per class. Represents the percentage of unknown samples on the total number of samples of that class seen during the stream.

    Returns
    -------
    dict
        Dictionary containing the unknown rate of each class
    """
    rates = {}
    for key, val in self.nb_class_unknown.items():
        total = self.class_sample_counter.get(key, 0)
        rates[key] = (val / total) if total > 0 else 0.0
    return rates

get_unknown_rate()

Returns the unknown rate, represents the percentage of unknown samples on the total number of samples classified in the online phase.

Returns:

Type Description
float

Unknown rate
Source code in streamndr/model/noveltydetectionclassifier.py 
18
19
20
21
22
23
24
25
26
27
28
def get_unknown_rate(self):
    """Returns the unknown rate, represents the percentage of unknown samples on the total number of samples classified in the online phase.

    Returns
    -------
    float
        Unknown rate
    """
    if self.sample_counter == 0:
        return 0.0
    return len(self.short_mem) / self.sample_counter