def random_splittable_value(self, variable: int) -> Any:
"""
Return a random value of a variable
Useful for choosing a variable to split on
Parameters
----------
variable - str
Name of the variable to split on
Returns
-------
Any
Notes
-----
- Won't create degenerate splits, all splits will have at least one row on both sides of the split
"""
if variable not in self.splittable_variables():
raise NoSplittableVariableException()
max_value = self.max_value_of_column(variable)
candidate = np.random.choice(self.get_column(variable))
while candidate == max_value:
candidate = np.random.choice(self.get_column(variable))
return candidate
def random_splittable_leaf_node(tree: Tree) -> LeafNode:
"""
Returns a random leaf node that can be split in a non-degenerate way
i.e. a random draw from the set of leaf nodes that have at least two distinct values in their covariate matrix
"""
splittable_nodes = tree.splittable_leaf_nodes
if len(splittable_nodes) == 0:
raise NoSplittableVariableException()
else:
return np.random.choice(splittable_nodes)
def random_splittable_variable(self) -> str:
"""
Choose a variable at random from the set of splittable variables
Returns
-------
str - a variable name that can be split on
"""
splittable_variables = list(self.splittable_variables())
if len(splittable_variables) == 0:
raise NoSplittableVariableException()
return np.random.choice(np.array(list(splittable_variables)), 1)[0]
def random_splittable_variable(self) -> str:
"""
Choose a variable at random from the set of splittable variables
Returns
-------
str - a variable name that can be split on
"""
if self.is_at_least_one_splittable_variable():
return np.random.choice(np.array(self.splittable_variables()),
1)[0]
else:
raise NoSplittableVariableException()