scikit hierarchical refactoring pseudo code

gistfile1.py

Dendogram = namedtuple('Dendogram', ['children', 'n_leaves', 'n_components'])


def hierarchical_tree(X, linkage_criterion='ward', connectivity=None, n_components=None, copy=True):
    ...

    for ...:


        if linkage_criterion == 'ward':
            # merge base on inertia criteria
        if linkage_criterion == 'complete':
            # merge base on complete criteria

    children = np.array(children)  # return numpy array for efficient caching
    dendogram = Dendogram(children, n_components, n_leaves)
    return dendogram


class HierarchicalClustering(BaseEstimator):
    def __init__(self, linkage_criterion='ward', metric='euclidean', n_clusters=2,
                 memory=Memory(cachedir=None, verbose=0),
                 connectivity=None, copy=True, n_components=None):
        self.linkage_criterion = linkage_criterion
        self.metric = metric
        self.n_clusters = n_clusters
        self.memory = memory
        self.copy = copy
        self.n_components = n_components
        self.connectivity = connectivity

    def fit(X):
        if iscallable(linkage_criterion):
            self.dendogram_ = linkage_criterion(X, ...)
        else:
            if self.connectivity is None:
                out = hierarchy.linkage(X, method=self.linkage_criterion, metric=self.metric)
                children = out[:, :2].astype(np.int)
                self.dendogram_ = Dendogram(children, 1, X.shape[0])
            else:
                self.dendogram_ = hierarchical_tree(X, ...)
        return self