def indicator_matrix(df):
''' The indicator matrix of the original dataframe. '''
indicator_matrix = pd.get_dummies(df)
for column in df.columns:
if df[column].dtype in ('int64', 'float64'):
indicator_matrix[column] = util.rescale(df[column], new_min=0, new_max=1)
return indicator_matrix
def indicator_matrix(df):
"""The indicator matrix of the original dataframe."""
indicator_matrix = pd.get_dummies(df)
for column in df.columns:
if df[column].dtype in ('int64', 'float64'):
indicator_matrix[column] = util.rescale(df[column], new_min=0, new_max=1)
return indicator_matrix
def _build_indicator_matrix(self):
''' Build the indicator matrix by placing a `1` where a row takes a value for a variable and
a `0` when it doesn't. '''
indicator_matrix = pd.get_dummies(self.X)
# Add the numerical variables after rescaling
for col in self.numerical_variables.columns:
indicator_matrix[col] = util.rescale(self.numerical_variables[col], new_min=0,
new_max=1)
return indicator_matrix
def _compute(self):
# Build the indicator matrix
self.indicator_matrix = pd.get_dummies(self.df)
# Add the numerical variables after rescaling
for col in self._numerical.columns:
self.indicator_matrix[col] = util.rescale(self._numerical[col])
# Compute the correspondance analysis of the indicator matrix
ca = prince.CA(self.indicator_matrix, components=self.components)
# Extract the needed values
self.column_projections = ca.column_projections
self.row_projections = ca.row_projections
self.U = ca.U
self.W = ca.W
self.V = ca.V
# Compute Benzécri scores instead of classical eigenvalues
self.eigenvalues = [
((self.Q / (self.Q - 1)) ** 2 * (w - 1 / self.Q) ** 2) ** 2
if w < 1 / self.Q
else w ** 2
for w in self.W
]
self._compute_explained_variance()
