Method skewness

pattern/vector/__init__.py:1732–1741

Yields 0.0 if the trained classes are evenly distributed. Yields > +1.0 or < -1.0 if the training data is highly skewed.

(self)

        
    @property
    def skewness(self):
        """ Yields 0.0 if the trained classes are evenly distributed.
            Yields > +1.0 or < -1.0 if the training data is highly skewed.
        """
        def moment(a, m, k=1):
            return sum([(x-m)**k for x in a]) / (len(a) or 1)
        # List each training instance by an int that represents its class:
        a = list(chain(*([i] * v for i, (k, v) in enumerate(self._classes.iteritems()))))
        m = float(sum(a)) / len(a) # mean
        return moment(a, m, 3) / (moment(a, m, 2) ** 1.5 or 1)

    def train(self, document, type=None):
        """ Trains the classifier with the given document of the given type (i.e., class).