def test_get_distribution(self):
d = data.Table("iris")
cls = d.domain.class_var
disc = distribution.get_distribution(d, cls)
self.assertIsInstance(disc, np.ndarray)
self.assertIs(disc.variable, cls)
self.assertEqual(disc.unknowns, 0)
np.testing.assert_array_equal(disc, [50, 50, 50])
petal_length = d.columns.petal_length
freqs = np.array(
[
(1.0, 1),
(1.1, 1),
(1.2, 2),
(1.3, 7),
(1.4, 12),
(1.5, 14),
(1.6, 7),
(1.7, 4),
(1.9, 2),
(3.0, 1),
(3.3, 2),
(3.5, 2),
(3.6, 1),
(3.7, 1),
(3.8, 1),
(3.9, 3),
(4.0, 5),
(4.1, 3),
(4.2, 4),
(4.3, 2),
(4.4, 4),
(4.5, 8),
(4.6, 3),
(4.7, 5),
(4.8, 4),
(4.9, 5),
(5.0, 4),
(5.1, 8),
(5.2, 2),
(5.3, 2),
(5.4, 2),
(5.5, 3),
(5.6, 6),
(5.7, 3),
(5.8, 3),
(5.9, 2),
(6.0, 2),
(6.1, 3),
(6.3, 1),
(6.4, 1),
(6.6, 1),
(6.7, 2),
(6.9, 1),
]
).T
disc = distribution.get_distribution(d, petal_length)
np.testing.assert_almost_equal(disc, freqs)
def column_imputer_random(variable, data):
if variable.is_discrete:
dist = distribution.get_distribution(data, variable)
transformer = RandomTransform(variable, dist)
elif variable.is_continuous:
dist = distribution.get_distribution(data, variable)
transformer = RandomTransform(variable, dist)
return RandomImputerModel((variable,), (variable,), (transformer,))
def column_imputer_random(variable, data):
if isinstance(variable, Orange.data.DiscreteVariable):
dist = distribution.get_distribution(data, variable)
transformer = RandomTransform(variable, dist)
elif isinstance(variable, Orange.data.ContinuousVariable):
dist = distribution.get_distribution(data, variable)
transformer = RandomTransform(variable, dist)
return RandomImputerModel((variable,), (variable,), (transformer,))
def compute_box_data(self):
attr = self.attribute
if not attr:
return
dataset = self.dataset
self.is_continuous = attr.is_continuous
if dataset is None or not self.is_continuous and not attr.values or \
self.group_var and not self.group_var.values:
self.stats = self.dist = self.conts = []
return
if self.group_var:
self.dist = []
self.conts = contingency.get_contingency(
dataset, attr, self.group_var)
if self.is_continuous:
self.stats = [BoxData(cont, attr, i, self.group_var)
for i, cont in enumerate(self.conts)]
self.label_txts_all = self.group_var.values
else:
self.dist = distribution.get_distribution(dataset, attr)
self.conts = []
if self.is_continuous:
self.stats = [BoxData(self.dist, attr, None)]
self.label_txts_all = [""]
self.label_txts = [txts for stat, txts in zip(self.stats,
self.label_txts_all)
if stat.n > 0]
self.stats = [stat for stat in self.stats if stat.n > 0]
def _setup(self):
self.plot.clear()
self.plot_prob.clear()
self._legend.clear()
self._legend.hide()
varidx = self.variable_idx
self.var = self.cvar = None
if varidx >= 0:
self.var = self.varmodel[varidx]
if self.groupvar_idx > 0:
self.cvar = self.groupvarmodel[self.groupvar_idx]
self.cb_prob.clear()
self.cb_prob.addItem("(None)")
self.cb_prob.addItems(self.cvar.values)
self.cb_prob.addItem("(All)")
self.show_prob = min(max(self.show_prob, 0), len(self.cvar.values) + 1)
data = self.data
self._setup_smoothing()
if self.var is None:
return
if self.disc_cont:
data = self.data[:, (self.var, self.cvar) if self.cvar else self.var]
disc = Orange.preprocess.discretize.EqualWidth(n=self.bins[self.smoothing_index])
data = Orange.preprocess.Discretize(data, method=disc, remove_const=False)
self.var = data.domain[0]
self.set_left_axis_name()
self.enable_disable_rel_freq()
if self.cvar:
self.contingencies = contingency.get_contingency(data, self.var, self.cvar)
self.display_contingency()
else:
self.distributions = distribution.get_distribution(data, self.var)
self.display_distribution()
self.plot.autoRange()
def _setup(self):
self.plot.clear()
self.plot_prob.clear()
self._legend.clear()
self._legend.hide()
varidx = self.variable_idx
self.var = self.cvar = None
if varidx >= 0:
self.var = self.varmodel[varidx]
if self.groupvar_idx > 0:
self.cvar = self.groupvarmodel[self.groupvar_idx]
data = self.data
self._setup_smoothing()
if self.var is None:
return
if self.disc_cont:
data = self.data[:, (self.var, self.cvar) if self.cvar else self.var ]
disc = Orange.preprocess.discretize.EqualWidth(n=self.bins[self.smoothing_index])
data = Orange.preprocess.Discretize(data, method=disc)
self.var = data.domain.variables[0]
self.set_left_axis_name()
self.enable_disable_rel_freq()
if self.cvar:
self.contingencies = \
contingency.get_contingency(data, self.var, self.cvar)
self.display_contingency()
else:
self.distributions = \
distribution.get_distribution(data, self.var)
self.display_distribution()
self.plot.autoRange()
def fit_storage(self, data):
dist = distribution.get_distribution(data, data.domain.class_var)
N = dist.sum()
if N > 0:
dist /= N
else:
dist.fill(1 / len(dist))
return ConstantClassifier(dist=dist)
def compute_box_data(self):
if self.split_var:
return (
contingency.get_contingency(
self.dataset, self.attribute, self.split_var),
self.split_var.values)
else:
return [
distribution.get_distribution(
self.dataset, self.attribute)], [""]
def fit_storage(self, dat):
if not isinstance(dat.domain.class_var, data.DiscreteVariable):
raise ValueError("classification.MajorityFitter expects a domain with a "
"(single) discrete variable")
dist = distribution.get_distribution(dat, dat.domain.class_var)
N = dist.sum()
if N > 0:
dist /= N
else:
dist.fill(1 / len(dist))
return ConstantClassifier(dist=dist)
def fit_storage(self, dat):
if not dat.domain.has_discrete_class:
raise ValueError("classification.MajorityLearner expects a domain with a "
"(single) discrete variable")
dist = distribution.get_distribution(dat, dat.domain.class_var)
N = dist.sum()
if N > 0:
dist /= N
else:
dist.fill(1 / len(dist))
return ConstantModel(dist=dist)
def __call__(self, data, attribute):
if type(data) == SqlTable:
att = attribute.to_sql()
quantiles = [(i + 1) / self.n for i in range(self.n - 1)]
query = data._sql_query(['quantile(%s, ARRAY%s)' %
(att, str(quantiles))])
with data._execute_sql_query(query) as cur:
points = sorted(set(cur.fetchone()[0]))
else:
d = distribution.get_distribution(data, attribute)
points = _discretize.split_eq_freq(d, n=self.n)
return Discretizer.create_discretized_var(
data.domain[attribute], points)
def _ensure_dist(var, data_or_dist):
if isinstance(data_or_dist, distribution.Discrete):
if not is_discrete(var):
raise TypeError
return data_or_dist
elif isinstance(data_or_dist, distribution.Continuous):
if not is_continuous(var):
raise TypeError
return data_or_dist
elif isinstance(data_or_dist, Orange.data.Storage):
return distribution.get_distribution(data_or_dist, var)
else:
raise ValueError("Need a distribution or data.")
def __call__(self, data, attribute):
if type(data) == SqlTable:
att = attribute.to_sql()
quantiles = [(i + 1) / self.n for i in range(self.n - 1)]
query = data._sql_query(
['quantile(%s, ARRAY%s)' % (att, str(quantiles))])
with data._execute_sql_query(query) as cur:
points = sorted(set(cur.fetchone()[0]))
else:
d = distribution.get_distribution(data, attribute)
points = _discretize.split_eq_freq(d, self.n)
return Discretizer.create_discretized_var(data.domain[attribute],
points)
def __call__(self, data, variable):
if is_continuous(variable):
stats = basic_stats.BasicStats(data, variable)
value = stats.mean
elif is_discrete(variable):
dist = distribution.get_distribution(data, variable)
value = dist.modus()
else:
raise TypeError
var = copy.copy(variable)
var.compute_value = ReplaceUnknowns(variable, value)
return var
def _ensure_dist(var, data_or_dist):
if isinstance(data_or_dist, distribution.Discrete):
if not var.is_discrete:
raise TypeError
return data_or_dist
elif isinstance(data_or_dist, distribution.Continuous):
if not var.is_continuous:
raise TypeError
return data_or_dist
elif isinstance(data_or_dist, Orange.data.Storage):
return distribution.get_distribution(data_or_dist, var)
else:
raise ValueError("Need a distribution or data.")
def fit_storage(self, dat):
if not isinstance(dat.domain.class_var, data.DiscreteVariable):
raise ValueError(
"classification.MajorityLearner expects a domain with a "
"(single) discrete variable")
dist = distribution.get_distribution(dat, dat.domain.class_var)
N = dist.sum()
if N > 0:
dist /= N
else:
dist.fill(1 / len(dist))
return ConstantModel(dist=dist)
def __call__(self, data, variable, value=None):
variable = data.domain[variable]
if value is None:
if variable.is_continuous:
stats = basic_stats.BasicStats(data, variable)
value = stats.mean
elif variable.is_discrete:
dist = distribution.get_distribution(data, variable)
value = dist.modus()
else:
raise TypeError("Variable must be continuous or discrete")
return variable.copy(compute_value=ReplaceUnknowns(variable, value))
def fit_storage(self, data):
"""
Construct a :obj:`MeanModel` by computing the mean value of the given
data.
:param data: data table
:type data: Orange.data.Table
:return: regression model, which always returns mean value
:rtype: :obj:`MeanModel`
"""
if not data.domain.has_continuous_class:
raise ValueError("regression.MeanLearner expects a domain with a "
"(single) continuous variable")
dist = distribution.get_distribution(data, data.domain.class_var)
return MeanModel(dist)
def __call__(self, data, variable, value=None):
variable = data.domain[variable]
if value is None:
if is_continuous(variable):
stats = basic_stats.BasicStats(data, variable)
value = stats.mean
elif is_discrete(variable):
dist = distribution.get_distribution(data, variable)
value = dist.modus()
else:
raise TypeError("Variable must be continuous or discrete")
var = copy.copy(variable)
var.compute_value = ReplaceUnknowns(variable, value)
return var
def __call__(self, data, variable, value=None):
variable = data.domain[variable]
if value is None:
if variable.is_continuous:
stats = basic_stats.BasicStats(data, variable)
value = stats.mean
elif variable.is_discrete:
dist = distribution.get_distribution(data, variable)
value = dist.modus()
else:
raise TypeError("Variable must be numeric or categorical.")
a = variable.copy(compute_value=ReplaceUnknowns(variable, value))
a.to_sql = ImputeSql(variable, value)
return a
def __call__(self, data, attribute):
if type(data) == SqlTable:
att = attribute.to_sql()
quantiles = [(i + 1) / self.n for i in range(self.n - 1)]
query = data._sql_query(
['quantile(%s, ARRAY%s)' % (att, str(quantiles))],
use_time_sample=1000)
with data._execute_sql_query(query) as cur:
points = sorted(set(cur.fetchone()[0]))
else:
d = distribution.get_distribution(data, attribute)
points = _discretize.split_eq_freq(d, self.n)
# np.unique handles cases in which differences are below precision
points = list(np.unique(points))
return Discretizer.create_discretized_var(data.domain[attribute],
points)
def transform(var):
dist = distribution.get_distribution(data, var)
if self.center:
c = self.center(dist)
dist[0, :] -= c
else:
c = 0
if self.scale:
s = self.scale(dist)
if s < 1e-15:
s = 1
else:
s = 1
factor = 1 / s
return var.copy(compute_value=transformation.Normalizer(var, c, factor))
def compute_box_data(self):
attr = self.attribute
if not attr:
return
dataset = self.dataset
if dataset is None \
or not attr.is_continuous and not attr.values \
or self.group_var and not self.group_var.values:
self.stats = []
self.dist = self.conts = None
return
if self.group_var:
self.dist = None
missing_val_str = f"missing '{self.group_var.name}'"
group_var_labels = self.group_var.values + ("", )
if self.attribute.is_continuous:
stats, label_texts = [], []
attr_col = dataset.get_column_view(attr)[0].astype(float)
for group, value in \
zip(self._group_cols(dataset, self.group_var, attr_col),
group_var_labels):
if group.size:
stats.append(BoxData(group, value))
label_texts.append(value or missing_val_str)
self.stats = stats
self.label_txts_all = label_texts
else:
self.conts = contingency.get_contingency(
dataset, attr, self.group_var)
self.label_txts_all = [
v or missing_val_str for v, c in zip(
group_var_labels, self.conts.array_with_unknowns)
if np.sum(c) > 0
]
else:
self.conts = None
if self.attribute.is_continuous:
attr_col = dataset.get_column_view(attr)[0].astype(float)
self.stats = [BoxData(attr_col)]
else:
self.dist = distribution.get_distribution(dataset, attr)
self.label_txts_all = [""]
self.label_txts = [
txts for stat, txts in zip(self.stats, self.label_txts_all)
if stat.n > 0
]
self.stats = [stat for stat in self.stats if stat.n > 0]
def __call__(self, data, variable):
variable = data.domain[variable]
dist = distribution.get_distribution(data, variable)
# A distribution is invalid if a continuous variable's column does not
# contain any known values or if a discrete variable's .values == []
isinvalid = dist.size == 0
if isinvalid and variable.is_discrete:
assert len(variable.values) == 0
raise ValueError("'{}' has no values".format(variable))
elif isinvalid and variable.is_continuous:
raise ValueError("'{}' has an unknown distribution".format(variable))
if variable.is_discrete and numpy.sum(dist) == 0:
dist += 1 / len(dist)
elif variable.is_continuous and numpy.sum(dist[1, :]) == 0:
dist[1, :] += 1 / dist.shape[1]
return variable.copy(compute_value=ReplaceUnknownsRandom(variable, dist))
def _disc_plot(self):
var = self.var
self.ploti.getAxis("bottom").setTicks([list(enumerate(var.values))])
colors = [QColor(0, 128, 255)]
dist = distribution.get_distribution(self.data, self.var)
for i, freq in enumerate(dist):
tooltip = \
"<p style='white-space:pre;'>" \
f"<b>{escape(var.values[i])}</b>: {int(freq)} " \
f"({100 * freq / len(self.valid_data):.2f} %) "
self._add_bar(i - 0.5,
1,
0.1, [freq],
colors,
stacked=False,
expanded=False,
tooltip=tooltip)
def fit_storage(self, dat):
if not dat.domain.has_discrete_class:
raise ValueError("classification.MajorityLearner expects a domain " "with a (single) discrete variable")
dist = distribution.get_distribution(dat, dat.domain.class_var)
N = dist.sum()
if N > 0:
dist /= N
else:
dist.fill(1 / len(dist))
probs = np.array(dist)
ties = np.flatnonzero(probs == probs.max())
if len(ties) > 1:
random_idx = int(sha1(bytes(dat.Y)).hexdigest(), 16) % len(ties)
unif_maj = ties[random_idx]
else:
unif_maj = None
return ConstantModel(dist=dist, unif_maj=unif_maj)
def transform(var):
dist = distribution.get_distribution(data, var)
if self.center != self.NoCentering:
c = self.center(dist)
dist[0, :] -= c
else:
c = 0
if self.scale != self.NoScaling:
s = self.scale(dist)
if s < 1e-15:
s = 1
else:
s = 1
factor = 1 / s
transformed_var = var.copy(compute_value=transformation.Normalizer(var, c, factor))
if s != 1:
transformed_var.number_of_decimals = 3
return transformed_var
def fit_storage(self, dat):
if not dat.domain.has_discrete_class:
raise ValueError("classification.MajorityLearner expects a domain "
"with a (single) discrete variable")
dist = distribution.get_distribution(dat, dat.domain.class_var)
N = dist.sum()
if N > 0:
dist /= N
else:
dist.fill(1 / len(dist))
probs = np.array(dist)
ties = np.flatnonzero(probs == probs.max())
if len(ties) > 1:
random_idx = int(sha1(bytes(dat.Y)).hexdigest(), 16) % len(ties)
unif_maj = ties[random_idx]
else:
unif_maj = None
return ConstantModel(dist=dist, unif_maj=unif_maj)
def compute_box_data(self):
attr = self.attribute
if not attr:
return
dataset = self.dataset
if dataset is None \
or not attr.is_continuous and not attr.values \
or self.group_var and not self.group_var.values:
self.stats = []
self.dist = self.conts = None
return
if self.group_var:
self.dist = None
self.conts = contingency.get_contingency(dataset, attr,
self.group_var)
missing_val_str = f"missing '{self.group_var.name}'"
group_var_labels = self.group_var.values + ("", )
if self.attribute.is_continuous:
stats, label_texts = [], []
for cont, value in zip(self.conts.array_with_unknowns,
group_var_labels):
if np.sum(cont[1]):
stats.append(BoxData(cont, value))
label_texts.append(value or missing_val_str)
self.stats = stats
self.label_txts_all = label_texts
else:
self.label_txts_all = [
v or missing_val_str for v, c in zip(
group_var_labels, self.conts.array_with_unknowns)
if np.sum(c) > 0
]
else:
self.dist = distribution.get_distribution(dataset, attr)
self.conts = None
if self.attribute.is_continuous:
self.stats = [BoxData(self.dist, None)]
self.label_txts_all = [""]
self.label_txts = [
txts for stat, txts in zip(self.stats, self.label_txts_all)
if stat.n > 0
]
self.stats = [stat for stat in self.stats if stat.n > 0]
def __call__(self, data, variable):
variable = data.domain[variable]
dist = distribution.get_distribution(data, variable)
# A distribution is invalid if a continuous variable's column does not
# contain any known values or if a discrete variable's .values == []
isinvalid = dist.size == 0
if isinvalid and variable.is_discrete:
assert len(variable.values) == 0
raise ValueError("'{}' has no values".format(variable))
elif isinvalid and variable.is_continuous:
raise ValueError(
"'{}' has an unknown distribution".format(variable))
if variable.is_discrete and np.sum(dist) == 0:
dist += 1 / len(dist)
elif variable.is_continuous and np.sum(dist[1, :]) == 0:
dist[1, :] += 1 / dist.shape[1]
return variable.copy(
compute_value=ReplaceUnknownsRandom(variable, dist))
def compute_box_data(self):
attr = self.attribute
if not attr:
return
dataset = self.dataset
self.is_continuous = attr.is_continuous
if dataset is None or not self.is_continuous and not attr.values or \
self.group_var and not self.group_var.values:
self.stats = self.dist = self.conts = []
return
if self.group_var:
self.dist = []
self.conts = contingency.get_contingency(dataset, attr,
self.group_var)
group_var_labels = self.group_var.values + [
f"missing '{self.group_var.name}'"
]
if self.is_continuous:
stats, label_texts = [], []
for i, cont in enumerate(self.conts.array_with_unknowns):
if np.sum(cont[1]):
stats.append(BoxData(cont, attr, i, self.group_var))
label_texts.append(group_var_labels[i])
self.stats = stats
self.label_txts_all = label_texts
else:
self.label_txts_all = [
v for v, c in zip(group_var_labels,
self.conts.array_with_unknowns)
if np.sum(c) > 0
]
else:
self.dist = distribution.get_distribution(dataset, attr)
self.conts = []
if self.is_continuous:
self.stats = [BoxData(self.dist, attr, None)]
self.label_txts_all = [""]
self.label_txts = [
txts for stat, txts in zip(self.stats, self.label_txts_all)
if stat.n > 0
]
self.stats = [stat for stat in self.stats if stat.n > 0]
def _setup(self):
self.plot.clear()
self.plot_prob.clear()
self._legend.clear()
self._legend.hide()
varidx = self.variable_idx
self.var = self.cvar = None
if varidx >= 0:
self.var = self.varmodel[varidx]
if self.groupvar_idx > 0:
self.cvar = self.groupvarmodel[self.groupvar_idx]
prob = self.controls.show_prob
prob.clear()
prob.addItem("(None)")
prob.addItems(self.cvar.values)
prob.addItem("(All)")
self.show_prob = min(max(self.show_prob, 0),
len(self.cvar.values) + 1)
data = self.data
self._setup_smoothing()
if self.var is None:
return
if self.disc_cont:
domain = Orange.data.Domain(
[self.var, self.cvar] if self.cvar else [self.var])
data = Orange.data.Table(domain, data)
disc = EqualWidth(n=self.bins[self.smoothing_index])
data = Discretize(method=disc, remove_const=False)(data)
self.var = data.domain[0]
self.set_left_axis_name()
self.enable_disable_rel_freq()
self.controls.cumulative_distr.setDisabled(not self.var.is_continuous)
if self.cvar:
self.contingencies = \
contingency.get_contingency(data, self.var, self.cvar)
self.display_contingency()
else:
self.distributions = \
distribution.get_distribution(data, self.var)
self.display_distribution()
self.plot.autoRange()
def _setup(self):
self.plot.clear()
self.plot_prob.clear()
self._legend.clear()
self._legend.hide()
varidx = self.variable_idx
self.var = self.cvar = None
if varidx >= 0:
self.var = self.varmodel[varidx]
if self.groupvar_idx > 0:
self.cvar = self.groupvarmodel[self.groupvar_idx]
prob = self.controls.show_prob
prob.clear()
prob.addItem("(None)")
prob.addItems(self.cvar.values)
prob.addItem("(All)")
self.show_prob = min(max(self.show_prob, 0),
len(self.cvar.values) + 1)
data = self.data
self._setup_smoothing()
if self.var is None:
return
if self.disc_cont:
domain = Orange.data.Domain(
[self.var, self.cvar] if self.cvar else [self.var])
data = Orange.data.Table(domain, data)
disc = EqualWidth(n=self.bins[self.smoothing_index])
data = Discretize(method=disc, remove_const=False)(data)
self.var = data.domain[0]
self.set_left_axis_name()
self.enable_disable_rel_freq()
self.controls.cumulative_distr.setDisabled(not self.var.is_continuous)
if self.cvar:
self.contingencies = \
contingency.get_contingency(data, self.var, self.cvar)
self.display_contingency()
else:
self.distributions = \
distribution.get_distribution(data, self.var)
self.display_distribution()
self.plot.autoRange()
def _setup(self):
self.plot.clear()
self.plot_prob.clear()
self._legend.clear()
self._legend.hide()
varidx = self.variable_idx
self.var = self.cvar = None
if varidx >= 0:
self.var = self.varmodel[varidx]
if self.groupvar_idx > 0:
self.cvar = self.groupvarmodel[self.groupvar_idx]
self.cb_prob.clear()
self.cb_prob.addItem("(None)")
self.cb_prob.addItems(self.cvar.values)
self.cb_prob.addItem("(All)")
self.show_prob = min(max(self.show_prob, 0),
len(self.cvar.values) + 1)
data = self.data
self._setup_smoothing()
if self.var is None:
return
if self.disc_cont:
data = self.data[:,
(self.var, self.cvar) if self.cvar else self.var]
disc = Orange.preprocess.discretize.EqualWidth(
n=self.bins[self.smoothing_index])
data = Orange.preprocess.Discretize(data,
method=disc,
remove_const=False)
self.var = data.domain[0]
self.set_left_axis_name()
self.enable_disable_rel_freq()
if self.cvar:
self.contingencies = \
contingency.get_contingency(data, self.var, self.cvar)
self.display_contingency()
else:
self.distributions = \
distribution.get_distribution(data, self.var)
self.display_distribution()
self.plot.autoRange()
def __call__(self, data, attribute, fixed=None):
if fixed:
min, max = fixed[attribute.name]
points = self._split_eq_width_fixed(min, max, n=self.n)
else:
if type(data) == SqlTable:
att = attribute.to_sql()
query = data._sql_query(['min(%s)::double precision' % att,
'max(%s)::double precision' % att])
with data._execute_sql_query(query) as cur:
min, max = cur.fetchone()
dif = (max - min) / self.n
points = [min + (i + 1) * dif for i in range(self.n - 1)]
else:
# TODO: why is the whole distribution computed instead of
# just min/max
d = distribution.get_distribution(data, attribute)
points = self._split_eq_width(d, n=self.n)
return Discretizer.create_discretized_var(
data.domain[attribute], points)
def _setup(self):
self.plot.clear()
varidx = self.variable_idx
self.var = self.cvar = None
if varidx >= 0:
self.var = self.varmodel[varidx]
if self.groupvar_idx > 0:
self.cvar = self.groupvarmodel[self.groupvar_idx]
self.set_left_axis_name()
self.enable_disable_rel_freq()
if self.var is None:
return
if self.cvar:
self.contingencies = \
contingency.get_contingency(self.data, self.var, self.cvar)
self.display_contingency()
else:
self.distributions = \
distribution.get_distribution(self.data, self.var)
self.display_distribution()
def _setup(self):
self.plot.clear()
varidx = self.variable_idx
self.var = self.cvar = None
if varidx >= 0:
self.var = self.varmodel[varidx]
if self.groupvar_idx > 0:
self.cvar = self.groupvarmodel[self.groupvar_idx]
self.set_left_axis_name()
self.enable_disable_rel_freq()
if self.var is None:
return
if self.cvar:
self.contingencies = \
contingency.get_contingency(self.data, self.var, self.cvar)
self.display_contingency()
else:
self.distributions = \
distribution.get_distribution(self.data, self.var)
self.display_distribution()
def transform(var):
dist = distribution.get_distribution(data, var)
if self.center != self.NoCentering:
c = self.center(dist)
dist[0, :] -= c
else:
c = 0
if self.scale != self.NoScaling:
s = self.scale(dist)
if s < 1e-15:
s = 1
else:
s = 1
factor = 1 / s
transformed_var = var.copy(
compute_value=transformation.Normalizer(var, c, factor))
if s != 1:
transformed_var.number_of_decimals = 3
return transformed_var
def _disc_plot(self):
var = self.var
dist = distribution.get_distribution(self.data, self.var)
dist = np.array(dist) # Distribution misbehaves in further operations
if self.sort_by_freq:
order = np.argsort(dist)[::-1]
else:
order = np.arange(len(dist))
ordered_values = np.array(var.values)[order]
self.ploti.getAxis("bottom").setTicks([list(enumerate(ordered_values))])
colors = [QColor(0, 128, 255)]
for i, freq, desc in zip(count(), dist[order], ordered_values):
tooltip = \
"<p style='white-space:pre;'>" \
f"<b>{escape(desc)}</b>: {int(freq)} " \
f"({100 * freq / len(self.valid_data):.2f} %) "
self._add_bar(
i - 0.5, 1, 0.1, [freq], colors,
stacked=False, expanded=False, tooltip=tooltip, desc=desc)
def _setup(self):
"""Setup the plot."""
self.plot.clear()
varidx = self.variable_idx
var = cvar = None
if varidx >= 0:
var = self.varmodel[varidx]
if self.groupvar_idx >= 0:
cvar = self.groupvarmodel[self.groupvar_idx]
if var is None:
return
if is_discrete(cvar):
cont = contingency.get_contingency(self.data, var, cvar)
self.set_contingency(cont, var, cvar)
else:
dist = distribution.get_distribution(self.data, var)
self.set_distribution(dist, var)
def fit_storage(self, dat):
"""
Constructs `Orange.classification.majority.ConstantClassifier` from given data.
:param dat: table of data
:type dat: Orange.data.Table
:return: classification model, which always returns majority value
:rtype: Orange.classification.majority.ConstantClassifier
"""
if not isinstance(dat.domain.class_var, data.DiscreteVariable):
raise ValueError(
"classification.MajorityFitter expects a domain with a "
"(single) discrete variable")
dist = distribution.get_distribution(dat, dat.domain.class_var)
N = dist.sum()
if N > 0:
dist /= N
else:
dist.fill(1 / len(dist))
return ConstantClassifier(dist=dist)
def _setup(self):
"""Setup the plot."""
self.plot.clear()
varidx = self.variable_idx
var = cvar = None
if varidx >= 0:
var = self.varmodel[varidx]
if self.groupvar_idx >= 0:
cvar = self.groupvarmodel[self.groupvar_idx]
if var is None:
return
if is_discrete(cvar):
cont = contingency.get_contingency(self.data, var, cvar)
self.set_contingency(cont, var, cvar)
else:
dist = distribution.get_distribution(self.data, var)
self.set_distribution(dist, var)
def iterate_states(self, state):
"""
Iterate through all combinations of attributes as ordered by Relief,
starting with a single attribute if Mosaic is colored by class
distributions, and two if by Pearson.
"""
# If we put initialization of `self.attrs` to `initialize`,
# `score_heuristic` would be run on every call to master's `set_data`.
master = self.master
data = master.discrete_data
min_attrs, max_attrs = self.attr_range()
if min_attrs > max_attrs:
return
if state is None: # on the first call, compute order
if self._compute_class_dists():
self.marginal = get_distribution(data, data.domain.class_var)
self.marginal.normalize()
state = list(range(min_attrs))
else:
self.marginal = get_distributions(data)
for dist in self.marginal:
dist.normalize()
state = list(range(min_attrs))
n_attrs = len(data.domain.attributes)
while True:
yield state
# Reset while running; just abort
if self.attr_ordering is None:
break
for up, _ in enumerate(state):
state[up] += 1
if up + 1 == len(state) or state[up] < state[up + 1]:
break
state[up] = up
if state[-1] == len(self.attr_ordering):
if len(state) < min(max_attrs, n_attrs):
state = list(range(len(state) + 1))
else:
break
def compute_box_data(self):
attr = self.attribute
if not attr:
return
dataset = self.dataset
self.is_continuous = attr.is_continuous
if dataset is None or not self.is_continuous and not attr.values or \
self.group_var and not self.group_var.values:
self.stats = self.dist = self.conts = []
return
if self.group_var:
self.dist = []
self.conts = contingency.get_contingency(
dataset, attr, self.group_var)
if self.is_continuous:
stats, label_texts = [], []
for i, cont in enumerate(self.conts):
if np.sum(cont[1]):
stats.append(BoxData(cont, attr, i, self.group_var))
label_texts.append(self.group_var.values[i])
self.stats = stats
self.label_txts_all = label_texts
else:
self.label_txts_all = \
[v for v, c in zip(self.group_var.values, self.conts)
if np.sum(c) > 0]
else:
self.dist = distribution.get_distribution(dataset, attr)
self.conts = []
if self.is_continuous:
self.stats = [BoxData(self.dist, attr, None)]
self.label_txts_all = [""]
self.label_txts = [txts for stat, txts in zip(self.stats,
self.label_txts_all)
if stat.n > 0]
self.stats = [stat for stat in self.stats if stat.n > 0]
def iterate_states(self, state):
"""
Iterate through all combinations of attributes as ordered by Relief,
starting with a single attribute if Mosaic is colored by class
distributions, and two if by Pearson.
"""
# If we put initialization of `self.attrs` to `initialize`,
# `score_heuristic` would be run on every call to master's `set_data`.
master = self.master
data = master.discrete_data
if state is None: # on the first call, compute order
if self._compute_class_dists():
self.marginal = get_distribution(data, data.domain.class_var)
self.marginal.normalize()
state = [0]
else:
self.marginal = get_distributions(data)
for dist in self.marginal:
dist.normalize()
state = [0, 1]
n_attrs = len(data.domain.attributes)
while True:
yield state
# Reset while running; just abort
if self.attr_ordering is None:
break
for up, _ in enumerate(state):
state[up] += 1
if up + 1 == len(state) or state[up] < state[up + 1]:
break
state[up] = up
if state[-1] == len(self.attr_ordering):
if len(state) < min(self.max_attrs, n_attrs):
state = list(range(len(state) + 1))
else:
break
def column_imputer_modus(variable, table):
stat = distribution.get_distribution(table, variable)
column_imputer_defaults(variable, table, stat.modus())
def fit_storage(self, dat):
if not isinstance(dat.domain.class_var, data.ContinuousVariable):
raise ValueError("regression.MeanFitter expects a domain with a "
"(single) continuous variable")
dist = distribution.get_distribution(dat, dat.domain.class_var)
return MeanModel(dist)
def add_rect(x0,
x1,
y0,
y1,
condition="",
used_attrs=[],
used_vals=[],
attr_vals=""):
area_index = len(self.areas)
if x0 == x1:
x1 += 1
if y0 == y1:
y1 += 1
# rectangles of width and height 1 are not shown - increase
if x1 - x0 + y1 - y0 == 2:
y1 += 1
if class_var and class_var.is_discrete:
colors = [QColor(*col) for col in class_var.colors]
else:
colors = None
def select_area(_, ev):
self.select_area(area_index, ev)
def rect(x, y, w, h, z, pen_color=None, brush_color=None, **args):
if pen_color is None:
return CanvasRectangle(self.canvas,
x,
y,
w,
h,
z=z,
onclick=select_area,
**args)
if brush_color is None:
brush_color = pen_color
return CanvasRectangle(self.canvas,
x,
y,
w,
h,
pen_color,
brush_color,
z=z,
onclick=select_area,
**args)
def line(x1, y1, x2, y2):
r = QGraphicsLineItem(x1, y1, x2, y2, None)
self.canvas.addItem(r)
r.setPen(QPen(Qt.white, 2))
r.setZValue(30)
outer_rect = rect(x0, y0, x1 - x0, y1 - y0, 30)
self.areas.append((used_attrs, used_vals, outer_rect))
if not conditionaldict[attr_vals]:
return
if self.interior_coloring == self.PEARSON:
s = sum(apriori_dists[0])
expected = s * reduce(
mul, (apriori_dists[i][used_vals[i]] / float(s)
for i in range(len(used_vals))))
actual = conditionaldict[attr_vals]
pearson = (actual - expected) / sqrt(expected)
if pearson == 0:
ind = 0
else:
ind = max(0, min(int(log(abs(pearson), 2)), 3))
color = [self.RED_COLORS, self.BLUE_COLORS][pearson > 0][ind]
rect(x0, y0, x1 - x0, y1 - y0, -20, color)
outer_rect.setToolTip(
condition + "<hr/>" + "Expected instances: %.1f<br>"
"Actual instances: %d<br>"
"Standardized (Pearson) residual: %.1f" %
(expected, conditionaldict[attr_vals], pearson))
else:
cls_values = get_variable_values_sorted(class_var)
prior = get_distribution(data, class_var.name)
total = 0
for i, value in enumerate(cls_values):
val = conditionaldict[attr_vals + "-" + value]
if val == 0:
continue
if i == len(cls_values) - 1:
v = y1 - y0 - total
else:
v = ((y1 - y0) * val) / conditionaldict[attr_vals]
rect(x0, y0 + total, x1 - x0, v, -20, colors[i])
total += v
if self.use_boxes and \
abs(x1 - x0) > bar_width and \
abs(y1 - y0) > bar_width:
total = 0
line(x0 + bar_width, y0, x0 + bar_width, y1)
n = sum(prior)
for i, (val, color) in enumerate(zip(prior, colors)):
if i == len(prior) - 1:
h = y1 - y0 - total
else:
h = (y1 - y0) * val / n
rect(x0, y0 + total, bar_width, h, 20, color)
total += h
if conditionalsubsetdict:
if conditionalsubsetdict[attr_vals]:
counts = [
conditionalsubsetdict[attr_vals + "-" + val]
for val in cls_values
]
if sum(counts) == 1:
rect(x0 - 2,
y0 - 2,
x1 - x0 + 5,
y1 - y0 + 5,
-550,
colors[counts.index(1)],
Qt.white,
penWidth=2,
penStyle=Qt.DashLine)
if self.subset_data is not None:
line(x1 - bar_width, y0, x1 - bar_width, y1)
total = 0
n = conditionalsubsetdict[attr_vals]
if n:
for i, (cls, color) in \
enumerate(zip(cls_values, colors)):
val = conditionalsubsetdict[attr_vals +
"-" + cls]
if val == 0:
continue
if i == len(prior) - 1:
v = y1 - y0 - total
else:
v = ((y1 - y0) * val) / n
rect(x1 - bar_width, y0 + total, bar_width,
v, 15, color)
total += v
actual = [
conditionaldict[attr_vals + "-" + cls_values[i]]
for i in range(len(prior))
]
n_actual = sum(actual)
if n_actual > 0:
apriori = [prior[key] for key in cls_values]
n_apriori = sum(apriori)
text = "<br/>".join(
"<b>%s</b>: %d / %.1f%% (Expected %.1f / %.1f%%)" %
(cls, act, 100.0 * act / n_actual,
apr / n_apriori * n_actual, 100.0 * apr / n_apriori)
for cls, act, apr in zip(cls_values, actual, apriori))
else:
text = ""
outer_rect.setToolTip("{}<hr>Instances: {}<br><br>{}".format(
condition, n_actual, text[:-4]))
def update_graph(self):
spacing = self.SPACING
bar_width = self.BAR_WIDTH
def draw_data(attr_list,
x0_x1,
y0_y1,
side,
condition,
total_attrs,
used_attrs=[],
used_vals=[],
attr_vals=""):
x0, x1 = x0_x1
y0, y1 = y0_y1
if conditionaldict[attr_vals] == 0:
add_rect(x0,
x1,
y0,
y1,
"",
used_attrs,
used_vals,
attr_vals=attr_vals)
# store coordinates for later drawing of labels
draw_text(side, attr_list[0], (x0, x1), (y0, y1), total_attrs,
used_attrs, used_vals, attr_vals)
return
attr = attr_list[0]
# how much smaller rectangles do we draw
edge = len(attr_list) * spacing
values = get_variable_values_sorted(data.domain[attr])
if side % 2:
values = values[::-1] # reverse names if necessary
if side % 2 == 0: # we are drawing on the x axis
# remove the space needed for separating different attr. values
whole = max(0, (x1 - x0) - edge * (len(values) - 1))
if whole == 0:
edge = (x1 - x0) / float(len(values) - 1)
else: # we are drawing on the y axis
whole = max(0, (y1 - y0) - edge * (len(values) - 1))
if whole == 0:
edge = (y1 - y0) / float(len(values) - 1)
if attr_vals == "":
counts = [conditionaldict[val] for val in values]
else:
counts = [
conditionaldict[attr_vals + "-" + val] for val in values
]
total = sum(counts)
# if we are visualizing the third attribute and the first attribute
# has the last value, we have to reverse the order in which the
# boxes will be drawn otherwise, if the last cell, nearest to the
# labels of the fourth attribute, is empty, we wouldn't be able to
# position the labels
valrange = list(range(len(values)))
if len(attr_list + used_attrs) == 4 and len(used_attrs) == 2:
attr1values = get_variable_values_sorted(
data.domain[used_attrs[0]])
if used_vals[0] == attr1values[-1]:
valrange = valrange[::-1]
for i in valrange:
start = i * edge + whole * float(sum(counts[:i]) / total)
end = i * edge + whole * float(sum(counts[:i + 1]) / total)
val = values[i]
htmlval = getHtmlCompatibleString(val)
if attr_vals != "":
newattrvals = attr_vals + "-" + val
else:
newattrvals = val
tooltip = condition + 4 * " " + attr + \
": <b>" + htmlval + "</b><br>"
attrs = used_attrs + [attr]
vals = used_vals + [val]
common_args = attrs, vals, newattrvals
if side % 2 == 0: # if we are moving horizontally
if len(attr_list) == 1:
add_rect(x0 + start, x0 + end, y0, y1, tooltip,
*common_args)
else:
draw_data(attr_list[1:], (x0 + start, x0 + end),
(y0, y1), side + 1, tooltip, total_attrs,
*common_args)
else:
if len(attr_list) == 1:
add_rect(x0, x1, y0 + start, y0 + end, tooltip,
*common_args)
else:
draw_data(attr_list[1:], (x0, x1),
(y0 + start, y0 + end), side + 1, tooltip,
total_attrs, *common_args)
draw_text(side, attr_list[0], (x0, x1), (y0, y1), total_attrs,
used_attrs, used_vals, attr_vals)
def draw_text(side, attr, x0_x1, y0_y1, total_attrs, used_attrs,
used_vals, attr_vals):
x0, x1 = x0_x1
y0, y1 = y0_y1
if side in drawn_sides:
return
# the text on the right will be drawn when we are processing
# visualization of the last value of the first attribute
if side == 3:
attr1values = \
get_variable_values_sorted(data.domain[used_attrs[0]])
if used_vals[0] != attr1values[-1]:
return
if not conditionaldict[attr_vals]:
if side not in draw_positions:
draw_positions[side] = (x0, x1, y0, y1)
return
else:
if side in draw_positions:
# restore the positions of attribute values and name
(x0, x1, y0, y1) = draw_positions[side]
drawn_sides.add(side)
values = get_variable_values_sorted(data.domain[attr])
if side % 2:
values = values[::-1]
spaces = spacing * (total_attrs - side) * (len(values) - 1)
width = x1 - x0 - spaces * (side % 2 == 0)
height = y1 - y0 - spaces * (side % 2 == 1)
# calculate position of first attribute
currpos = 0
if attr_vals == "":
counts = [conditionaldict.get(val, 1) for val in values]
else:
counts = [
conditionaldict.get(attr_vals + "-" + val, 1)
for val in values
]
total = sum(counts)
if total == 0:
counts = [1] * len(values)
total = sum(counts)
aligns = [
Qt.AlignTop | Qt.AlignHCenter, Qt.AlignRight | Qt.AlignVCenter,
Qt.AlignBottom | Qt.AlignHCenter,
Qt.AlignLeft | Qt.AlignVCenter
]
align = aligns[side]
for i in range(len(values)):
val = values[i]
perc = counts[i] / float(total)
if distributiondict[val] != 0:
if side == 0:
CanvasText(self.canvas, str(val),
x0 + currpos + width * 0.5 * perc,
y1 + self.ATTR_VAL_OFFSET, align)
elif side == 1:
CanvasText(self.canvas, str(val),
x0 - self.ATTR_VAL_OFFSET,
y0 + currpos + height * 0.5 * perc, align)
elif side == 2:
CanvasText(self.canvas, str(val),
x0 + currpos + width * perc * 0.5,
y0 - self.ATTR_VAL_OFFSET, align)
else:
CanvasText(self.canvas, str(val),
x1 + self.ATTR_VAL_OFFSET,
y0 + currpos + height * 0.5 * perc, align)
if side % 2 == 0:
currpos += perc * width + spacing * (total_attrs - side)
else:
currpos += perc * height + spacing * (total_attrs - side)
if side == 0:
CanvasText(self.canvas,
attr,
x0 + (x1 - x0) / 2,
y1 + self.ATTR_VAL_OFFSET + self.ATTR_NAME_OFFSET,
align,
bold=1)
elif side == 1:
CanvasText(self.canvas,
attr,
x0 - max_ylabel_w1 - self.ATTR_VAL_OFFSET,
y0 + (y1 - y0) / 2,
align,
bold=1,
vertical=True)
elif side == 2:
CanvasText(self.canvas,
attr,
x0 + (x1 - x0) / 2,
y0 - self.ATTR_VAL_OFFSET - self.ATTR_NAME_OFFSET,
align,
bold=1)
else:
CanvasText(self.canvas,
attr,
x1 + max_ylabel_w2 + self.ATTR_VAL_OFFSET,
y0 + (y1 - y0) / 2,
align,
bold=1,
vertical=True)
def add_rect(x0,
x1,
y0,
y1,
condition="",
used_attrs=[],
used_vals=[],
attr_vals=""):
area_index = len(self.areas)
if x0 == x1:
x1 += 1
if y0 == y1:
y1 += 1
# rectangles of width and height 1 are not shown - increase
if x1 - x0 + y1 - y0 == 2:
y1 += 1
if class_var and class_var.is_discrete:
colors = [QColor(*col) for col in class_var.colors]
else:
colors = None
def select_area(_, ev):
self.select_area(area_index, ev)
def rect(x, y, w, h, z, pen_color=None, brush_color=None, **args):
if pen_color is None:
return CanvasRectangle(self.canvas,
x,
y,
w,
h,
z=z,
onclick=select_area,
**args)
if brush_color is None:
brush_color = pen_color
return CanvasRectangle(self.canvas,
x,
y,
w,
h,
pen_color,
brush_color,
z=z,
onclick=select_area,
**args)
def line(x1, y1, x2, y2):
r = QGraphicsLineItem(x1, y1, x2, y2, None)
self.canvas.addItem(r)
r.setPen(QPen(Qt.white, 2))
r.setZValue(30)
outer_rect = rect(x0, y0, x1 - x0, y1 - y0, 30)
self.areas.append((used_attrs, used_vals, outer_rect))
if not conditionaldict[attr_vals]:
return
if self.interior_coloring == self.PEARSON:
s = sum(apriori_dists[0])
expected = s * reduce(
mul, (apriori_dists[i][used_vals[i]] / float(s)
for i in range(len(used_vals))))
actual = conditionaldict[attr_vals]
pearson = (actual - expected) / sqrt(expected)
if pearson == 0:
ind = 0
else:
ind = max(0, min(int(log(abs(pearson), 2)), 3))
color = [self.RED_COLORS, self.BLUE_COLORS][pearson > 0][ind]
rect(x0, y0, x1 - x0, y1 - y0, -20, color)
outer_rect.setToolTip(
condition + "<hr/>" + "Expected instances: %.1f<br>"
"Actual instances: %d<br>"
"Standardized (Pearson) residual: %.1f" %
(expected, conditionaldict[attr_vals], pearson))
else:
cls_values = get_variable_values_sorted(class_var)
prior = get_distribution(data, class_var.name)
total = 0
for i, value in enumerate(cls_values):
val = conditionaldict[attr_vals + "-" + value]
if val == 0:
continue
if i == len(cls_values) - 1:
v = y1 - y0 - total
else:
v = ((y1 - y0) * val) / conditionaldict[attr_vals]
rect(x0, y0 + total, x1 - x0, v, -20, colors[i])
total += v
if self.use_boxes and \
abs(x1 - x0) > bar_width and \
abs(y1 - y0) > bar_width:
total = 0
line(x0 + bar_width, y0, x0 + bar_width, y1)
n = sum(prior)
for i, (val, color) in enumerate(zip(prior, colors)):
if i == len(prior) - 1:
h = y1 - y0 - total
else:
h = (y1 - y0) * val / n
rect(x0, y0 + total, bar_width, h, 20, color)
total += h
if conditionalsubsetdict:
if conditionalsubsetdict[attr_vals]:
counts = [
conditionalsubsetdict[attr_vals + "-" + val]
for val in cls_values
]
if sum(counts) == 1:
rect(x0 - 2,
y0 - 2,
x1 - x0 + 5,
y1 - y0 + 5,
-550,
colors[counts.index(1)],
Qt.white,
penWidth=2,
penStyle=Qt.DashLine)
if self.subset_data is not None:
line(x1 - bar_width, y0, x1 - bar_width, y1)
total = 0
n = conditionalsubsetdict[attr_vals]
if n:
for i, (cls, color) in \
enumerate(zip(cls_values, colors)):
val = conditionalsubsetdict[attr_vals +
"-" + cls]
if val == 0:
continue
if i == len(prior) - 1:
v = y1 - y0 - total
else:
v = ((y1 - y0) * val) / n
rect(x1 - bar_width, y0 + total, bar_width,
v, 15, color)
total += v
actual = [
conditionaldict[attr_vals + "-" + cls_values[i]]
for i in range(len(prior))
]
n_actual = sum(actual)
if n_actual > 0:
apriori = [prior[key] for key in cls_values]
n_apriori = sum(apriori)
text = "<br/>".join(
"<b>%s</b>: %d / %.1f%% (Expected %.1f / %.1f%%)" %
(cls, act, 100.0 * act / n_actual,
apr / n_apriori * n_actual, 100.0 * apr / n_apriori)
for cls, act, apr in zip(cls_values, actual, apriori))
else:
text = ""
outer_rect.setToolTip("{}<hr>Instances: {}<br><br>{}".format(
condition, n_actual, text[:-4]))
def draw_legend(x0_x1, y0_y1):
x0, x1 = x0_x1
y0, y1 = y0_y1
if self.interior_coloring == self.PEARSON:
names = [
"<-8", "-8:-4", "-4:-2", "-2:2", "2:4", "4:8", ">8",
"Residuals:"
]
colors = self.RED_COLORS[::-1] + self.BLUE_COLORS[1:]
else:
names = get_variable_values_sorted(class_var) + \
[class_var.name + ":"]
colors = [QColor(*col) for col in class_var.colors]
names = [
CanvasText(self.canvas, name, alignment=Qt.AlignVCenter)
for name in names
]
totalwidth = sum(text.boundingRect().width() for text in names)
# compute the x position of the center of the legend
y = y1 + self.ATTR_NAME_OFFSET + self.ATTR_VAL_OFFSET + 35
distance = 30
startx = (x0 + x1) / 2 - (totalwidth + (len(names)) * distance) / 2
names[-1].setPos(startx + 15, y)
names[-1].show()
xoffset = names[-1].boundingRect().width() + distance
size = 8
for i in range(len(names) - 1):
if self.interior_coloring == self.PEARSON:
edgecolor = Qt.black
else:
edgecolor = colors[i]
CanvasRectangle(self.canvas, startx + xoffset, y - size / 2,
size, size, edgecolor, colors[i])
names[i].setPos(startx + xoffset + 10, y)
xoffset += distance + names[i].boundingRect().width()
self.canvas.clear()
self.areas = []
data = self.discrete_data
if data is None:
return
subset = self.subset_data
attr_list = self.get_attr_list()
class_var = data.domain.class_var
if class_var:
sql = type(data) == SqlTable
name = not sql and data.name
# save class_var because it is removed in the next line
data = data[:, attr_list + [class_var]]
data.domain.class_var = class_var
if not sql:
data.name = name
else:
data = data[:, attr_list]
# TODO: check this
# data = Preprocessor_dropMissing(data)
if len(data) == 0:
self.warning(5, "No valid data for current attributes.")
return
else:
self.warning(5)
if self.interior_coloring == self.PEARSON:
apriori_dists = [
get_distribution(data, attr) for attr in attr_list
]
else:
apriori_dists = []
def get_max_label_width(attr):
values = get_variable_values_sorted(data.domain[attr])
maxw = 0
for val in values:
t = CanvasText(self.canvas, val, 0, 0, bold=0, show=False)
maxw = max(int(t.boundingRect().width()), maxw)
return maxw
# get the maximum width of rectangle
xoff = 20
width = 20
if len(attr_list) > 1:
text = CanvasText(self.canvas, attr_list[1], bold=1, show=0)
max_ylabel_w1 = min(get_max_label_width(attr_list[1]), 150)
width = 5 + text.boundingRect().height() + \
self.ATTR_VAL_OFFSET + max_ylabel_w1
xoff = width
if len(attr_list) == 4:
text = CanvasText(self.canvas, attr_list[3], bold=1, show=0)
max_ylabel_w2 = min(get_max_label_width(attr_list[3]), 150)
width += text.boundingRect().height() + \
self.ATTR_VAL_OFFSET + max_ylabel_w2 - 10
# get the maximum height of rectangle
height = 100
yoff = 45
square_size = min(self.canvas_view.width() - width - 20,
self.canvas_view.height() - height - 20)
if square_size < 0:
return # canvas is too small to draw rectangles
self.canvas_view.setSceneRect(0, 0, self.canvas_view.width(),
self.canvas_view.height())
drawn_sides = set()
draw_positions = {}
conditionaldict, distributiondict = \
get_conditional_distribution(data, attr_list)
conditionalsubsetdict = None
if subset:
conditionalsubsetdict, _ = \
get_conditional_distribution(subset, attr_list)
# draw rectangles
draw_data(attr_list, (xoff, xoff + square_size),
(yoff, yoff + square_size), 0, "", len(attr_list))
draw_legend((xoff, xoff + square_size), (yoff, yoff + square_size))
self.update_selection_rects()
def update_graph(self):
spacing = self.SPACING
bar_width = self.BAR_WIDTH
def get_counts(attr_vals, values):
"""Calculate rectangles' widths; if all are 0, they are set to 1."""
if not attr_vals:
counts = [conditionaldict[val] for val in values]
else:
counts = [
conditionaldict[attr_vals + "-" + val] for val in values
]
total = sum(counts)
if total == 0:
counts = [1] * len(values)
total = sum(counts)
return total, counts
def draw_data(attr_list,
x0_x1,
y0_y1,
side,
condition,
total_attrs,
used_attrs,
used_vals,
attr_vals=""):
x0, x1 = x0_x1
y0, y1 = y0_y1
if conditionaldict[attr_vals] == 0:
add_rect(x0,
x1,
y0,
y1,
"",
used_attrs,
used_vals,
attr_vals=attr_vals)
# store coordinates for later drawing of labels
draw_text(side, attr_list[0], (x0, x1), (y0, y1), total_attrs,
used_attrs, used_vals, attr_vals)
return
attr = attr_list[0]
# how much smaller rectangles do we draw
edge = len(attr_list) * spacing
values = get_variable_values_sorted(attr)
if side % 2:
values = values[::-1] # reverse names if necessary
if side % 2 == 0: # we are drawing on the x axis
# remove the space needed for separating different attr. values
whole = max(0, (x1 - x0) - edge * (len(values) - 1))
if whole == 0:
edge = (x1 - x0) / float(len(values) - 1)
else: # we are drawing on the y axis
whole = max(0, (y1 - y0) - edge * (len(values) - 1))
if whole == 0:
edge = (y1 - y0) / float(len(values) - 1)
total, counts = get_counts(attr_vals, values)
# when visualizing the third attribute and the first attribute has
# the last value, reverse the order in which the boxes are drawn;
# otherwise, if the last cell, nearest to the labels of the fourth
# attribute, is empty, we wouldn't be able to position the labels
valrange = list(range(len(values)))
if len(attr_list + used_attrs) == 4 and len(used_attrs) == 2:
attr1values = get_variable_values_sorted(used_attrs[0])
if used_vals[0] == attr1values[-1]:
valrange = valrange[::-1]
for i in valrange:
start = i * edge + whole * float(sum(counts[:i]) / total)
end = i * edge + whole * float(sum(counts[:i + 1]) / total)
val = values[i]
htmlval = to_html(val)
newattrvals = attr_vals + "-" + val if attr_vals else val
tooltip = "{} {}: <b>{}</b><br/>".format(
condition, attr.name, htmlval)
attrs = used_attrs + [attr]
vals = used_vals + [val]
args = attrs, vals, newattrvals
if side % 2 == 0: # if we are moving horizontally
if len(attr_list) == 1:
add_rect(x0 + start, x0 + end, y0, y1, tooltip, *args)
else:
draw_data(attr_list[1:], (x0 + start, x0 + end),
(y0, y1), side + 1, tooltip, total_attrs,
*args)
else:
if len(attr_list) == 1:
add_rect(x0, x1, y0 + start, y0 + end, tooltip, *args)
else:
draw_data(attr_list[1:], (x0, x1),
(y0 + start, y0 + end), side + 1, tooltip,
total_attrs, *args)
draw_text(side, attr_list[0], (x0, x1), (y0, y1), total_attrs,
used_attrs, used_vals, attr_vals)
def draw_text(side, attr, x0_x1, y0_y1, total_attrs, used_attrs,
used_vals, attr_vals):
x0, x1 = x0_x1
y0, y1 = y0_y1
if side in drawn_sides:
return
# the text on the right will be drawn when we are processing
# visualization of the last value of the first attribute
if side == 3:
attr1values = get_variable_values_sorted(used_attrs[0])
if used_vals[0] != attr1values[-1]:
return
if not conditionaldict[attr_vals]:
if side not in draw_positions:
draw_positions[side] = (x0, x1, y0, y1)
return
else:
if side in draw_positions:
# restore the positions of attribute values and name
(x0, x1, y0, y1) = draw_positions[side]
drawn_sides.add(side)
values = get_variable_values_sorted(attr)
if side % 2:
values = values[::-1]
spaces = spacing * (total_attrs - side) * (len(values) - 1)
width = x1 - x0 - spaces * (side % 2 == 0)
height = y1 - y0 - spaces * (side % 2 == 1)
# calculate position of first attribute
currpos = 0
total, counts = get_counts(attr_vals, values)
aligns = [
Qt.AlignTop | Qt.AlignHCenter, Qt.AlignRight | Qt.AlignVCenter,
Qt.AlignBottom | Qt.AlignHCenter,
Qt.AlignLeft | Qt.AlignVCenter
]
align = aligns[side]
for i, val in enumerate(values):
if distributiondict[val] != 0:
perc = counts[i] / float(total)
rwidth = width * perc
xs = [
x0 + currpos + rwidth / 2, x0 - self.ATTR_VAL_OFFSET,
x0 + currpos + rwidth / 2, x1 + self.ATTR_VAL_OFFSET
]
ys = [
y1 + self.ATTR_VAL_OFFSET,
y0 + currpos + height * 0.5 * perc,
y0 - self.ATTR_VAL_OFFSET,
y0 + currpos + height * 0.5 * perc
]
CanvasText(self.canvas,
val,
xs[side],
ys[side],
align,
max_width=rwidth if side == 0 else None)
space = height if side % 2 else width
currpos += perc * space + spacing * (total_attrs - side)
xs = [
x0 + (x1 - x0) / 2, x0 - max_ylabel_w1 - self.ATTR_VAL_OFFSET,
x0 + (x1 - x0) / 2, x1 + max_ylabel_w2 + self.ATTR_VAL_OFFSET
]
ys = [
y1 + self.ATTR_VAL_OFFSET + self.ATTR_NAME_OFFSET,
y0 + (y1 - y0) / 2,
y0 - self.ATTR_VAL_OFFSET - self.ATTR_NAME_OFFSET,
y0 + (y1 - y0) / 2
]
CanvasText(self.canvas,
attr.name,
xs[side],
ys[side],
align,
bold=True,
vertical=side % 2)
def add_rect(x0,
x1,
y0,
y1,
condition,
used_attrs,
used_vals,
attr_vals=""):
area_index = len(self.areas)
x1 += (x0 == x1)
y1 += (y0 == y1)
# rectangles of width and height 1 are not shown - increase
y1 += (x1 - x0 + y1 - y0 == 2)
colors = class_var and [QColor(*col) for col in class_var.colors]
def select_area(_, ev):
self.select_area(area_index, ev)
def rect(x, y, w, h, z, pen_color=None, brush_color=None, **args):
if pen_color is None:
return CanvasRectangle(self.canvas,
x,
y,
w,
h,
z=z,
onclick=select_area,
**args)
if brush_color is None:
brush_color = pen_color
return CanvasRectangle(self.canvas,
x,
y,
w,
h,
pen_color,
brush_color,
z=z,
onclick=select_area,
**args)
def line(x1, y1, x2, y2):
r = QGraphicsLineItem(x1, y1, x2, y2, None)
self.canvas.addItem(r)
r.setPen(QPen(Qt.white, 2))
r.setZValue(30)
outer_rect = rect(x0, y0, x1 - x0, y1 - y0, 30)
self.areas.append((used_attrs, used_vals, outer_rect))
if not conditionaldict[attr_vals]:
return
if self.variable_color is None:
s = sum(apriori_dists[0])
expected = s * reduce(
mul, (apriori_dists[i][used_vals[i]] / float(s)
for i in range(len(used_vals))))
actual = conditionaldict[attr_vals]
pearson = float((actual - expected) / sqrt(expected))
if pearson == 0:
ind = 0
else:
ind = max(0, min(int(log(abs(pearson), 2)), 3))
color = [self.RED_COLORS, self.BLUE_COLORS][pearson > 0][ind]
rect(x0, y0, x1 - x0, y1 - y0, -20, color)
outer_rect.setToolTip(
condition + "<hr/>" + "Expected instances: %.1f<br>"
"Actual instances: %d<br>"
"Standardized (Pearson) residual: %.1f" %
(expected, conditionaldict[attr_vals], pearson))
else:
cls_values = get_variable_values_sorted(class_var)
prior = get_distribution(data, class_var.name)
total = 0
for i, value in enumerate(cls_values):
val = conditionaldict[attr_vals + "-" + value]
if val == 0:
continue
if i == len(cls_values) - 1:
v = y1 - y0 - total
else:
v = ((y1 - y0) * val) / conditionaldict[attr_vals]
rect(x0, y0 + total, x1 - x0, v, -20, colors[i])
total += v
if self.use_boxes and \
abs(x1 - x0) > bar_width and abs(y1 - y0) > bar_width:
total = 0
line(x0 + bar_width, y0, x0 + bar_width, y1)
n = sum(prior)
for i, (val, color) in enumerate(zip(prior, colors)):
if i == len(prior) - 1:
h = y1 - y0 - total
else:
h = (y1 - y0) * val / n
rect(x0, y0 + total, bar_width, h, 20, color)
total += h
if conditionalsubsetdict:
if conditionalsubsetdict[attr_vals]:
if self.subset_indices is not None:
line(x1 - bar_width, y0, x1 - bar_width, y1)
total = 0
n = conditionalsubsetdict[attr_vals]
if n:
for i, (cls, color) in \
enumerate(zip(cls_values, colors)):
val = conditionalsubsetdict[attr_vals +
"-" + cls]
if val == 0:
continue
if i == len(prior) - 1:
v = y1 - y0 - total
else:
v = ((y1 - y0) * val) / n
rect(x1 - bar_width, y0 + total, bar_width,
v, 15, color)
total += v
actual = [
conditionaldict[attr_vals + "-" + cls_values[i]]
for i in range(len(prior))
]
n_actual = sum(actual)
if n_actual > 0:
apriori = [prior[key] for key in cls_values]
n_apriori = sum(apriori)
text = "<br/>".join(
"<b>%s</b>: %d / %.1f%% (Expected %.1f / %.1f%%)" %
(cls, act, 100.0 * act / n_actual,
apr / n_apriori * n_actual, 100.0 * apr / n_apriori)
for cls, act, apr in zip(cls_values, actual, apriori))
else:
text = ""
outer_rect.setToolTip("{}<hr>Instances: {}<br><br>{}".format(
condition, n_actual, text[:-4]))
def create_legend():
if self.variable_color is None:
names = [
"<-8", "-8:-4", "-4:-2", "-2:2", "2:4", "4:8", ">8",
"Residuals:"
]
colors = self.RED_COLORS[::-1] + self.BLUE_COLORS[1:]
edges = repeat(Qt.black)
else:
names = get_variable_values_sorted(class_var)
edges = colors = [QColor(*col) for col in class_var.colors]
items = []
size = 8
for name, color, edgecolor in zip(names, colors, edges):
item = QGraphicsItemGroup()
item.addToGroup(
CanvasRectangle(None, -size / 2, -size / 2, size, size,
edgecolor, color))
item.addToGroup(
CanvasText(None, name, size, 0, Qt.AlignVCenter))
items.append(item)
return wrap_legend_items(items,
hspacing=20,
vspacing=16 + size,
max_width=self.canvas_view.width() - xoff)
self.canvas.clear()
self.areas = []
data = self.discrete_data
if data is None:
return
attr_list = self.get_disc_attr_list()
class_var = data.domain.class_var
# TODO: check this
# data = Preprocessor_dropMissing(data)
unique = [v.name for v in set(attr_list + [class_var]) if v]
if len(data[:, unique]) == 0:
self.Warning.no_valid_data()
return
else:
self.Warning.no_valid_data.clear()
attrs = [attr for attr in attr_list if not attr.values]
if attrs:
CanvasText(self.canvas,
"Feature {} has no values".format(attrs[0]),
(self.canvas_view.width() - 120) / 2,
self.canvas_view.height() / 2)
return
if self.variable_color is None:
apriori_dists = [
get_distribution(data, attr) for attr in attr_list
]
else:
apriori_dists = []
def get_max_label_width(attr):
values = get_variable_values_sorted(attr)
maxw = 0
for val in values:
t = CanvasText(self.canvas, val, 0, 0, bold=0, show=False)
maxw = max(int(t.boundingRect().width()), maxw)
return maxw
xoff = 20
# get the maximum width of rectangle
width = 20
max_ylabel_w1 = max_ylabel_w2 = 0
if len(attr_list) > 1:
text = CanvasText(self.canvas, attr_list[1].name, bold=1, show=0)
max_ylabel_w1 = min(get_max_label_width(attr_list[1]), 150)
width = 5 + text.boundingRect().height() + \
self.ATTR_VAL_OFFSET + max_ylabel_w1
xoff = width
if len(attr_list) == 4:
text = CanvasText(self.canvas,
attr_list[3].name,
bold=1,
show=0)
max_ylabel_w2 = min(get_max_label_width(attr_list[3]), 150)
width += text.boundingRect().height() + \
self.ATTR_VAL_OFFSET + max_ylabel_w2 - 10
legend = create_legend()
# get the maximum height of rectangle
yoff = 45
legendoff = yoff + self.ATTR_NAME_OFFSET + self.ATTR_VAL_OFFSET + 35
square_size = min(
self.canvas_view.width() - width - 20,
self.canvas_view.height() - legendoff -
legend.boundingRect().height())
if square_size < 0:
return # canvas is too small to draw rectangles
self.canvas_view.setSceneRect(0, 0, self.canvas_view.width(),
self.canvas_view.height())
drawn_sides = set()
draw_positions = {}
conditionaldict, distributiondict = \
get_conditional_distribution(data, attr_list)
conditionalsubsetdict = None
if self.subset_indices:
conditionalsubsetdict, _ = get_conditional_distribution(
self.discrete_data[self.subset_indices], attr_list)
# draw rectangles
draw_data(attr_list, (xoff, xoff + square_size),
(yoff, yoff + square_size), 0, "", len(attr_list), [], [])
self.canvas.addItem(legend)
legend.setPos(
xoff - legend.boundingRect().x() +
max(0, (square_size - legend.boundingRect().width()) / 2),
legendoff + square_size)
self.update_selection_rects()
def add_rect(x0, x1, y0, y1, condition="",
used_attrs=[], used_vals=[], attr_vals=""):
area_index = len(self.areas)
if x0 == x1:
x1 += 1
if y0 == y1:
y1 += 1
# rectangles of width and height 1 are not shown - increase
if x1 - x0 + y1 - y0 == 2:
y1 += 1
if class_var and class_var.is_discrete:
colors = [QColor(*col) for col in class_var.colors]
else:
colors = None
def select_area(_, ev):
self.select_area(area_index, ev)
def rect(x, y, w, h, z, pen_color=None, brush_color=None, **args):
if pen_color is None:
return CanvasRectangle(
self.canvas, x, y, w, h, z=z, onclick=select_area,
**args)
if brush_color is None:
brush_color = pen_color
return CanvasRectangle(
self.canvas, x, y, w, h, pen_color, brush_color, z=z,
onclick=select_area, **args)
def line(x1, y1, x2, y2):
r = QGraphicsLineItem(x1, y1, x2, y2, None)
self.canvas.addItem(r)
r.setPen(QPen(Qt.white, 2))
r.setZValue(30)
outer_rect = rect(x0, y0, x1 - x0, y1 - y0, 30)
self.areas.append((used_attrs, used_vals, outer_rect))
if not conditionaldict[attr_vals]:
return
if self.interior_coloring == self.PEARSON:
s = sum(apriori_dists[0])
expected = s * reduce(
mul,
(apriori_dists[i][used_vals[i]] / float(s)
for i in range(len(used_vals))))
actual = conditionaldict[attr_vals]
pearson = (actual - expected) / sqrt(expected)
if pearson == 0:
ind = 0
else:
ind = max(0, min(int(log(abs(pearson), 2)), 3))
color = [self.RED_COLORS, self.BLUE_COLORS][pearson > 0][ind]
rect(x0, y0, x1 - x0, y1 - y0, -20, color)
outer_rect.setToolTip(
condition + "<hr/>" +
"Expected instances: %.1f<br>"
"Actual instances: %d<br>"
"Standardized (Pearson) residual: %.1f" %
(expected, conditionaldict[attr_vals], pearson))
else:
cls_values = get_variable_values_sorted(class_var)
prior = get_distribution(data, class_var.name)
total = 0
for i, value in enumerate(cls_values):
val = conditionaldict[attr_vals + "-" + value]
if val == 0:
continue
if i == len(cls_values) - 1:
v = y1 - y0 - total
else:
v = ((y1 - y0) * val) / conditionaldict[attr_vals]
rect(x0, y0 + total, x1 - x0, v, -20, colors[i])
total += v
if self.use_boxes and \
abs(x1 - x0) > bar_width and \
abs(y1 - y0) > bar_width:
total = 0
line(x0 + bar_width, y0, x0 + bar_width, y1)
n = sum(prior)
for i, (val, color) in enumerate(zip(prior, colors)):
if i == len(prior) - 1:
h = y1 - y0 - total
else:
h = (y1 - y0) * val / n
rect(x0, y0 + total, bar_width, h, 20, color)
total += h
if conditionalsubsetdict:
if conditionalsubsetdict[attr_vals]:
counts = [conditionalsubsetdict[attr_vals + "-" + val]
for val in cls_values]
if sum(counts) == 1:
rect(x0 - 2, y0 - 2, x1 - x0 + 5, y1 - y0 + 5, -550,
colors[counts.index(1)], Qt.white,
penWidth=2, penStyle=Qt.DashLine)
if self.subset_data is not None:
line(x1 - bar_width, y0, x1 - bar_width, y1)
total = 0
n = conditionalsubsetdict[attr_vals]
if n:
for i, (cls, color) in \
enumerate(zip(cls_values, colors)):
val = conditionalsubsetdict[
attr_vals + "-" + cls]
if val == 0:
continue
if i == len(prior) - 1:
v = y1 - y0 - total
else:
v = ((y1 - y0) * val) / n
rect(x1 - bar_width, y0 + total,
bar_width, v, 15, color)
total += v
actual = [conditionaldict[attr_vals + "-" + cls_values[i]]
for i in range(len(prior))]
n_actual = sum(actual)
if n_actual > 0:
apriori = [prior[key] for key in cls_values]
n_apriori = sum(apriori)
text = "<br/>".join(
"<b>%s</b>: %d / %.1f%% (Expected %.1f / %.1f%%)" %
(cls, act, 100.0 * act / n_actual,
apr / n_apriori * n_actual, 100.0 * apr / n_apriori
)
for cls, act, apr in zip(cls_values, actual, apriori
))
else:
text = ""
outer_rect.setToolTip(
"{}<hr>Instances: {}<br><br>{}".format(
condition, n_actual, text[:-4]))
def fit_storage(self, data):
dist = distribution.get_distribution(data, data.domain.class_var)
return MeanModel(dist)
def fit_storage(self, data):
dist = distribution.get_distribution(data, data.domain.class_var)
domain = Orange.data.Domain((), (data.domain.class_var,))
return MeanPredictor(domain, dist)
def column_imputer_modus(variable, table):
stat = distribution.get_distribution(table, variable)
column_imputer_defaults(variable, table, stat.modus())
def fit_storage(self, data):
dist = distribution.get_distribution(data, data.domain.class_var)
domain = Orange.data.Domain((), (data.domain.class_var,))
return MeanPredictor(domain, dist)
def update_graph(self):
spacing = self.SPACING
bar_width = self.BAR_WIDTH
def draw_data(attr_list, x0_x1, y0_y1, side, condition,
total_attrs, used_attrs=[], used_vals=[],
attr_vals=""):
x0, x1 = x0_x1
y0, y1 = y0_y1
if conditionaldict[attr_vals] == 0:
add_rect(x0, x1, y0, y1, "",
used_attrs, used_vals, attr_vals=attr_vals)
# store coordinates for later drawing of labels
draw_text(side, attr_list[0], (x0, x1), (y0, y1), total_attrs,
used_attrs, used_vals, attr_vals)
return
attr = attr_list[0]
# how much smaller rectangles do we draw
edge = len(attr_list) * spacing
values = get_variable_values_sorted(data.domain[attr])
if side % 2:
values = values[::-1] # reverse names if necessary
if side % 2 == 0: # we are drawing on the x axis
# remove the space needed for separating different attr. values
whole = max(0, (x1 - x0) - edge * (
len(values) - 1))
if whole == 0:
edge = (x1 - x0) / float(len(values) - 1)
else: # we are drawing on the y axis
whole = max(0, (y1 - y0) - edge * (len(values) - 1))
if whole == 0:
edge = (y1 - y0) / float(len(values) - 1)
if attr_vals == "":
counts = [conditionaldict[val] for val in values]
else:
counts = [conditionaldict[attr_vals + "-" + val]
for val in values]
total = sum(counts)
# if we are visualizing the third attribute and the first attribute
# has the last value, we have to reverse the order in which the
# boxes will be drawn otherwise, if the last cell, nearest to the
# labels of the fourth attribute, is empty, we wouldn't be able to
# position the labels
valrange = list(range(len(values)))
if len(attr_list + used_attrs) == 4 and len(used_attrs) == 2:
attr1values = get_variable_values_sorted(
data.domain[used_attrs[0]])
if used_vals[0] == attr1values[-1]:
valrange = valrange[::-1]
for i in valrange:
start = i * edge + whole * float(sum(counts[:i]) / total)
end = i * edge + whole * float(sum(counts[:i + 1]) / total)
val = values[i]
htmlval = to_html(val)
if attr_vals != "":
newattrvals = attr_vals + "-" + val
else:
newattrvals = val
tooltip = condition + 4 * " " + attr + \
": <b>" + htmlval + "</b><br>"
attrs = used_attrs + [attr]
vals = used_vals + [val]
common_args = attrs, vals, newattrvals
if side % 2 == 0: # if we are moving horizontally
if len(attr_list) == 1:
add_rect(x0 + start, x0 + end, y0, y1,
tooltip, *common_args)
else:
draw_data(attr_list[1:], (x0 + start, x0 + end),
(y0, y1), side + 1,
tooltip, total_attrs, *common_args)
else:
if len(attr_list) == 1:
add_rect(x0, x1, y0 + start, y0 + end,
tooltip, *common_args)
else:
draw_data(attr_list[1:], (x0, x1),
(y0 + start, y0 + end), side + 1,
tooltip, total_attrs, *common_args)
draw_text(side, attr_list[0], (x0, x1), (y0, y1),
total_attrs, used_attrs, used_vals, attr_vals)
def draw_text(side, attr, x0_x1, y0_y1,
total_attrs, used_attrs, used_vals, attr_vals):
x0, x1 = x0_x1
y0, y1 = y0_y1
if side in drawn_sides:
return
# the text on the right will be drawn when we are processing
# visualization of the last value of the first attribute
if side == 3:
attr1values = \
get_variable_values_sorted(data.domain[used_attrs[0]])
if used_vals[0] != attr1values[-1]:
return
if not conditionaldict[attr_vals]:
if side not in draw_positions:
draw_positions[side] = (x0, x1, y0, y1)
return
else:
if side in draw_positions:
# restore the positions of attribute values and name
(x0, x1, y0, y1) = draw_positions[side]
drawn_sides.add(side)
values = get_variable_values_sorted(data.domain[attr])
if side % 2:
values = values[::-1]
spaces = spacing * (total_attrs - side) * (len(values) - 1)
width = x1 - x0 - spaces * (side % 2 == 0)
height = y1 - y0 - spaces * (side % 2 == 1)
# calculate position of first attribute
currpos = 0
if attr_vals == "":
counts = [conditionaldict.get(val, 1) for val in values]
else:
counts = [conditionaldict.get(attr_vals + "-" + val, 1)
for val in values]
total = sum(counts)
if total == 0:
counts = [1] * len(values)
total = sum(counts)
aligns = [Qt.AlignTop | Qt.AlignHCenter,
Qt.AlignRight | Qt.AlignVCenter,
Qt.AlignBottom | Qt.AlignHCenter,
Qt.AlignLeft | Qt.AlignVCenter]
align = aligns[side]
for i in range(len(values)):
val = values[i]
perc = counts[i] / float(total)
if distributiondict[val] != 0:
if side == 0:
CanvasText(self.canvas, str(val),
x0 + currpos + width * 0.5 * perc,
y1 + self.ATTR_VAL_OFFSET, align)
elif side == 1:
CanvasText(self.canvas, str(val),
x0 - self.ATTR_VAL_OFFSET,
y0 + currpos + height * 0.5 * perc, align)
elif side == 2:
CanvasText(self.canvas, str(val),
x0 + currpos + width * perc * 0.5,
y0 - self.ATTR_VAL_OFFSET, align)
else:
CanvasText(self.canvas, str(val),
x1 + self.ATTR_VAL_OFFSET,
y0 + currpos + height * 0.5 * perc, align)
if side % 2 == 0:
currpos += perc * width + spacing * (total_attrs - side)
else:
currpos += perc * height + spacing * (total_attrs - side)
if side == 0:
CanvasText(
self.canvas, attr,
x0 + (x1 - x0) / 2,
y1 + self.ATTR_VAL_OFFSET +
self.ATTR_NAME_OFFSET,
align, bold=1)
elif side == 1:
CanvasText(
self.canvas, attr,
x0 - max_ylabel_w1 - self.ATTR_VAL_OFFSET,
y0 + (y1 - y0) / 2,
align, bold=1, vertical=True)
elif side == 2:
CanvasText(
self.canvas, attr,
x0 + (x1 - x0) / 2,
y0 - self.ATTR_VAL_OFFSET -
self.ATTR_NAME_OFFSET,
align, bold=1)
else:
CanvasText(
self.canvas, attr,
x1 + max_ylabel_w2 + self.ATTR_VAL_OFFSET,
y0 + (y1 - y0) / 2,
align, bold=1, vertical=True)
def add_rect(x0, x1, y0, y1, condition="",
used_attrs=[], used_vals=[], attr_vals=""):
area_index = len(self.areas)
if x0 == x1:
x1 += 1
if y0 == y1:
y1 += 1
# rectangles of width and height 1 are not shown - increase
if x1 - x0 + y1 - y0 == 2:
y1 += 1
if class_var and class_var.is_discrete:
colors = [QColor(*col) for col in class_var.colors]
else:
colors = None
def select_area(_, ev):
self.select_area(area_index, ev)
def rect(x, y, w, h, z, pen_color=None, brush_color=None, **args):
if pen_color is None:
return CanvasRectangle(
self.canvas, x, y, w, h, z=z, onclick=select_area,
**args)
if brush_color is None:
brush_color = pen_color
return CanvasRectangle(
self.canvas, x, y, w, h, pen_color, brush_color, z=z,
onclick=select_area, **args)
def line(x1, y1, x2, y2):
r = QGraphicsLineItem(x1, y1, x2, y2, None)
self.canvas.addItem(r)
r.setPen(QPen(Qt.white, 2))
r.setZValue(30)
outer_rect = rect(x0, y0, x1 - x0, y1 - y0, 30)
self.areas.append((used_attrs, used_vals, outer_rect))
if not conditionaldict[attr_vals]:
return
if self.interior_coloring == self.PEARSON:
s = sum(apriori_dists[0])
expected = s * reduce(
mul,
(apriori_dists[i][used_vals[i]] / float(s)
for i in range(len(used_vals))))
actual = conditionaldict[attr_vals]
pearson = (actual - expected) / sqrt(expected)
if pearson == 0:
ind = 0
else:
ind = max(0, min(int(log(abs(pearson), 2)), 3))
color = [self.RED_COLORS, self.BLUE_COLORS][pearson > 0][ind]
rect(x0, y0, x1 - x0, y1 - y0, -20, color)
outer_rect.setToolTip(
condition + "<hr/>" +
"Expected instances: %.1f<br>"
"Actual instances: %d<br>"
"Standardized (Pearson) residual: %.1f" %
(expected, conditionaldict[attr_vals], pearson))
else:
cls_values = get_variable_values_sorted(class_var)
prior = get_distribution(data, class_var.name)
total = 0
for i, value in enumerate(cls_values):
val = conditionaldict[attr_vals + "-" + value]
if val == 0:
continue
if i == len(cls_values) - 1:
v = y1 - y0 - total
else:
v = ((y1 - y0) * val) / conditionaldict[attr_vals]
rect(x0, y0 + total, x1 - x0, v, -20, colors[i])
total += v
if self.use_boxes and \
abs(x1 - x0) > bar_width and \
abs(y1 - y0) > bar_width:
total = 0
line(x0 + bar_width, y0, x0 + bar_width, y1)
n = sum(prior)
for i, (val, color) in enumerate(zip(prior, colors)):
if i == len(prior) - 1:
h = y1 - y0 - total
else:
h = (y1 - y0) * val / n
rect(x0, y0 + total, bar_width, h, 20, color)
total += h
if conditionalsubsetdict:
if conditionalsubsetdict[attr_vals]:
counts = [conditionalsubsetdict[attr_vals + "-" + val]
for val in cls_values]
if sum(counts) == 1:
rect(x0 - 2, y0 - 2, x1 - x0 + 5, y1 - y0 + 5, -550,
colors[counts.index(1)], Qt.white,
penWidth=2, penStyle=Qt.DashLine)
if self.subset_data is not None:
line(x1 - bar_width, y0, x1 - bar_width, y1)
total = 0
n = conditionalsubsetdict[attr_vals]
if n:
for i, (cls, color) in \
enumerate(zip(cls_values, colors)):
val = conditionalsubsetdict[
attr_vals + "-" + cls]
if val == 0:
continue
if i == len(prior) - 1:
v = y1 - y0 - total
else:
v = ((y1 - y0) * val) / n
rect(x1 - bar_width, y0 + total,
bar_width, v, 15, color)
total += v
actual = [conditionaldict[attr_vals + "-" + cls_values[i]]
for i in range(len(prior))]
n_actual = sum(actual)
if n_actual > 0:
apriori = [prior[key] for key in cls_values]
n_apriori = sum(apriori)
text = "<br/>".join(
"<b>%s</b>: %d / %.1f%% (Expected %.1f / %.1f%%)" %
(cls, act, 100.0 * act / n_actual,
apr / n_apriori * n_actual, 100.0 * apr / n_apriori
)
for cls, act, apr in zip(cls_values, actual, apriori
))
else:
text = ""
outer_rect.setToolTip(
"{}<hr>Instances: {}<br><br>{}".format(
condition, n_actual, text[:-4]))
def draw_legend(x0_x1, y0_y1):
x0, x1 = x0_x1
y0, y1 = y0_y1
if self.interior_coloring == self.PEARSON:
names = ["<-8", "-8:-4", "-4:-2", "-2:2", "2:4", "4:8", ">8",
"Residuals:"]
colors = self.RED_COLORS[::-1] + self.BLUE_COLORS[1:]
else:
names = get_variable_values_sorted(class_var) + \
[class_var.name + ":"]
colors = [QColor(*col) for col in class_var.colors]
names = [CanvasText(self.canvas, name, alignment=Qt.AlignVCenter)
for name in names]
totalwidth = sum(text.boundingRect().width() for text in names)
# compute the x position of the center of the legend
y = y1 + self.ATTR_NAME_OFFSET + self.ATTR_VAL_OFFSET + 35
distance = 30
startx = (x0 + x1) / 2 - (totalwidth + (len(names)) * distance) / 2
names[-1].setPos(startx + 15, y)
names[-1].show()
xoffset = names[-1].boundingRect().width() + distance
size = 8
for i in range(len(names) - 1):
if self.interior_coloring == self.PEARSON:
edgecolor = Qt.black
else:
edgecolor = colors[i]
CanvasRectangle(self.canvas, startx + xoffset, y - size / 2,
size, size, edgecolor, colors[i])
names[i].setPos(startx + xoffset + 10, y)
xoffset += distance + names[i].boundingRect().width()
self.canvas.clear()
self.areas = []
data = self.discrete_data
if data is None:
return
subset = self.subset_data
attr_list = self.get_attr_list()
class_var = data.domain.class_var
if class_var:
sql = type(data) == SqlTable
name = not sql and data.name
# save class_var because it is removed in the next line
data = data[:, attr_list + [class_var]]
data.domain.class_var = class_var
if not sql:
data.name = name
else:
data = data[:, attr_list]
# TODO: check this
# data = Preprocessor_dropMissing(data)
if len(data) == 0:
self.Warning.no_valid_data()
return
else:
self.Warning.no_valid_data.clear()
if self.interior_coloring == self.PEARSON:
apriori_dists = [get_distribution(data, attr) for attr in attr_list]
else:
apriori_dists = []
def get_max_label_width(attr):
values = get_variable_values_sorted(data.domain[attr])
maxw = 0
for val in values:
t = CanvasText(self.canvas, val, 0, 0, bold=0, show=False)
maxw = max(int(t.boundingRect().width()), maxw)
return maxw
# get the maximum width of rectangle
xoff = 20
width = 20
if len(attr_list) > 1:
text = CanvasText(self.canvas, attr_list[1], bold=1, show=0)
max_ylabel_w1 = min(get_max_label_width(attr_list[1]), 150)
width = 5 + text.boundingRect().height() + \
self.ATTR_VAL_OFFSET + max_ylabel_w1
xoff = width
if len(attr_list) == 4:
text = CanvasText(self.canvas, attr_list[3], bold=1, show=0)
max_ylabel_w2 = min(get_max_label_width(attr_list[3]), 150)
width += text.boundingRect().height() + \
self.ATTR_VAL_OFFSET + max_ylabel_w2 - 10
# get the maximum height of rectangle
height = 100
yoff = 45
square_size = min(self.canvas_view.width() - width - 20,
self.canvas_view.height() - height - 20)
if square_size < 0:
return # canvas is too small to draw rectangles
self.canvas_view.setSceneRect(
0, 0, self.canvas_view.width(), self.canvas_view.height())
drawn_sides = set()
draw_positions = {}
conditionaldict, distributiondict = \
get_conditional_distribution(data, attr_list)
conditionalsubsetdict = None
if subset:
conditionalsubsetdict, _ = \
get_conditional_distribution(subset, attr_list)
# draw rectangles
draw_data(
attr_list, (xoff, xoff + square_size), (yoff, yoff + square_size),
0, "", len(attr_list))
draw_legend((xoff, xoff + square_size), (yoff, yoff + square_size))
self.update_selection_rects()