def analyze_last_to_move_all(graphs, nodes, weight='weight'):
last = graphs[-1]
observed = _crosstab(last, nodes, weight)
ca = CA()
ca.fit(observed)
_project(last, ca, observed)
for g in graphs[:-1]:
if sorted(g.nodes) != sorted(last.nodes):
raise ValueError('all graphs must have the same nodes')
observed = _crosstab(g, nodes, weight)
_project(g, ca, observed)
def plot_ca(self, groupby, len_set=None, labels_in=None):
ca = CA(n_components=2, n_iter=3, random_state=101)
df = self.get_union(groupby=groupby,
len_set=len_set,
labels_in=labels_in)
ca.fit(df)
ax = ca.plot_coordinates(X=df, figsize=(20, 8))
ax.get_legend().remove()
plt.savefig(
"%s%s%s对应分析图.png" % (self.savepath, self.name, groupby),
format="png",
bbox_inches="tight",
transparent=True,
dpi=600,
)
def setup_class(cls):
# Load a dataframe
cls.initial_dataframe = pd.read_csv('tests/data/presidentielles07.csv', index_col=0)
# Detemine the shape of the initial dataframe
(cls.n, cls.p) = cls.initial_dataframe.shape
# Determine the total number of observations
cls.N = np.sum(cls.initial_dataframe.values)
# Calculate a full CA
cls.n_components = cls.p
cls.ca = CA(cls.initial_dataframe, n_components=cls.n_components)
def corplot_graph(g, nodes, weight='weight', plot=True):
other = [m for m in g.nodes if m not in nodes]
for n, m in g.edges:
if (n in nodes and m in nodes) or (n in other and m in other):
raise ValueError('nodes do not define a bipartition')
if isinstance(g, nx.DiGraph) and (n in other and m in nodes):
raise ValueError('nodes do not define a directed bipartition')
nodes = [n for n in nodes if g.degree(n) > 0]
other = [m for m in other if g.degree(m) > 0]
sparse = nx.bipartite.biadjacency_matrix(g, nodes, other, weight=weight)
matrix = sparse.toarray()
observed = pd.DataFrame(matrix, nodes, other)
ca = CA()
ca.fit(observed)
if plot:
ca.plot_coordinates(observed)
h = g.subgraph(nodes + other).copy()
pos = ca.row_coordinates(observed)
for n, x, y in zip(nodes, pos[0], pos[1]):
h.nodes[n]['pos'] = (x, y)
h.nodes[n]['color'] = (76, 116, 172)
pos = ca.column_coordinates(observed)
for m, x, y in zip(other, pos[0], pos[1]):
h.nodes[m]['pos'] = (x, y)
h.nodes[m]['color'] = (219, 130, 87)
normalize(h)
return h
def analyze_to_move(g, nodes, weight='weight'):
observed = _crosstab(g, nodes, weight)
ca = CA()
ca.fit(observed)
_project(g, ca, observed)
def corplot_twomode(g, nodes, weight='weight'):
observed = _crosstab(g, nodes, weight)
ca = CA()
ca.fit(observed)
ca.plot_coordinates(observed)
def corplot_loose(x, y):
observed = pd.crosstab(_series(x), _series(y))
ca = CA()
ca.fit(observed)
ca.plot_coordinates(observed)
def ca(df, k):
"""The executed CA."""
return CA(df, n_components=k)
def corplot(df, x, y):
observed = pd.crosstab(df[y], df[x])
ca = CA()
ca.fit(observed)
ca.plot_coordinates(observed)