Tree recursion on python

Question

I have the following tree algorithm which prints the conditions for each leaf:

def _grow_tree(self, X, y, depth=0):
 # Identify best split
 idx, thr = self._best_split(X, y)
 # Indentation for tree description
 indent = " " * depth
 indices_left = X.iloc[:, idx] < thr
 X_left = X[indices_left]
 y_left = y_train[X_left.reset_index().loc[:,'id'].values]
 X_right = X[~indices_left]
 y_right = y_train[X_right.reset_index().loc[:,'id'].values]
 self.tree_describe.append(indent +"if x['"+ X.columns[idx] + "'] <= " +\
 str(thr) + ':')
 # Grow on left side of the tree 
 node.left = self._grow_tree(X_left, y_left, depth + 1)
 self.tree_describe.append(indent +"else: #if x['"+ X.columns[idx] + "'] > " +\
 str(thr) + ':')
 # Grow on right side of the tree
 node.right = self._grow_tree(X_right, y_right, depth + 1)
 return node

This produces the following print for a particular case:

["if x['VAR1'] <= 0.5:",
 " if x['VAR2'] <= 0.5:",
 " else: #if x['VAR2'] > 0.5:",
 "else: #if x['VAR1'] > 0.5:",
 " if x['VAR3'] <= 0.5:",
 " else: #if x['VAR3'] > 0.5:"]

How could I obtain the following output?:

["if x['VAR1'] <= 0.5:",
 " if x['VAR1'] <= 0.5&x['VAR2'] <= 0.5",
 " else: #if x['VAR1'] <= 0.5&x['VAR2'] > 0.5:",
 "else: #if x['VAR1'] > 0.5:",
 " if x['VAR1'] > 0.5&x['VAR3'] <= 0.5:",
 " else: #if x['VAR1'] > 0.5&x['VAR3'] > 0.5:"]

• h4z3: Didn't you make the indentation because you didn't want this output? Because now indent shows which item is a child of another, and now you want to repeat the x['VAR1'] <= 0.5/x['VAR1'] > 0.5 parts to show that.
• Jojo: Initially yes, but now I intend to use the pandas query function to create leaf columns based on conditions and the first way is not practical. I need for each leaf to have all the conditions.
• h4z3: So just forward it like you did with depth. Depth grows +1, description elements will grow with conditions.

Answer 1

You could introduce a new argument to your function, which will have the string with higher-level condition(s) that need to be added to each deeper conditions:

I would also suggest using .format() for your string building:

def _grow_tree(self, X, y, depth=0, descr=""):
 idx, thr = self._best_split(X, y)
 indent = " " * depth
 cond = "x['{}'] <= {}{}".format(X.columns[idx], thr, descr)
 self.tree_describe.append("{}if {}:".format(indent, cond))
 node.left = self._grow_tree(X_left, y_left, depth + 1, " & " + cond)
 cond = "x['{}'] > {}{}".format(X.columns[idx], thr, descr)
 self.tree_describe.append("{}else: #if {}:".format(indent, cond))
 node.right = self._grow_tree(X_right, y_right, depth + 1, " & " + cond)
 return node