from dataclasses import dataclass
from typing import Literal
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from utils import *
from metrics import *
np.random.seed(42)
@dataclass
class Next:
def __init__(self,name, ptr = None):
self.ptr = ptr
self.name = name
class Node:
def __init__(self, to_split=None, split_val=0.0, depth=float('inf'), decision = None):
self.to_split = to_split
self.split_val = split_val
self.depth = depth
self.decision = decision
self.left_ptr = None
self.right_ptr = None
self.nexts = []
def add_in_nexts(self,name, ptr=None):
lst = self.nexts
lst.append(Next(name = name,ptr = ptr))
self.nexts = lst
class DecisionTree:
criterion: Literal["information_gain", "gini_index"]
max_depth: int
def __init__(self, criterion='information_gain', max_depth=4):
self.criterion = criterion
self.max_depth = max_depth
self.Root = None
def fit(self, X: pd.DataFrame, y: pd.Series) -> None:
"""
Function to train and construct the decision tree
"""
def dfconvertor(a):
x=pd.DataFrame(a[0])
x[a[1].name]=a[1]
x=x.reset_index(drop=True)
return x.iloc[:,:-1], x.iloc[:,-1]
def make_tree(X: pd.DataFrame, y:pd.Series, features:list, criterion:str, max_depth:int, node_, value):
# same condition for all cases
if(y.nunique() == 1):
node_.decision = y.mode()[0]
return
# for discrete outputs
elif(len(features) == 0 and (not check_ifreal(y))):
node_.decision = y.mode()[0]
return
elif (node_.depth > max_depth and (not check_ifreal(y))):
node_.decision = y.mode()[0]
return
# for real outputs
elif(len(features) == 0 and (check_ifreal(y))):
node_.decision = y.mean()
return
elif (node_.depth > max_depth and (check_ifreal(y))):
node_.decision = y.mean()
return
else:
# for discrete inputs
if (not check_ifreal(X.iloc[:,0])):
opt_attr_ = opt_split_attribute(X, y, criterion, pd.Series(features))
node_.to_split = opt_attr_
for element in split_data(X, y, opt_attr_, 0):
Xi, yi = dfconvertor(element)
node_.add_in_nexts(name = Xi[opt_attr_].unique()[0])
new_node = Node(depth = node_.depth + 1, split_val=value)
node_.nexts[-1].ptr = new_node
features_ = [feature for feature in features if feature != opt_attr_]
make_tree(Xi, yi, features_, criterion, max_depth, new_node, value)
return
# for real inputs
else:
# for RIRO
if check_ifreal(y):
opt_attribute, split_val = opt_attr(X, y)
#for RIDO
else:
optimal_attribute = find_optimal_attribute(df_to_array(X), np.array(y))
opt_attribute = X.columns[optimal_attribute.feature_index]
split_val = optimal_attribute.split_value
# for real input
node_.to_split = opt_attribute
node_.split_val = split_val
t, r = split_data(X, y, opt_attribute, split_val)
left_df_X, left_df_y = dfconvertor(t)
right_df_X, right_df_y = dfconvertor(r)
new_node = Node(depth=node_.depth + 1)
node_.left_ptr = new_node
# features_ = [feature for feature in features if feature != opt_attribute]
make_tree(left_df_X, left_df_y, features, criterion, max_depth, new_node, value)
new_node_ = Node(depth=node_.depth + 1)
node_.right_ptr = new_node_
# _features_ = [feature for feature in features if feature != opt_attribute]
make_tree(right_df_X, right_df_y, features, criterion, max_depth, new_node_, value)
return
self.Root = Node(depth = 0, split_val = y.mean() if check_ifreal(y) else y.mode()[0])
make_tree(X, y, list(X.columns), self.criterion, self.max_depth, self.Root, self.Root.split_val)
return None
pass
def predict(self, X: pd.DataFrame) -> pd.Series:
"""
Funtion to run the decision tree on test inputs
"""
# for discrete inputs
# if (not check_ifreal(X.iloc[:,0])):
# def predict_from_node(feature_names: list, feature_values: list, node):
# if (node.decision is not None):
# return node.decision
# else:
# split_attribute = node.to_split
# val = feature_values[feature_names.index(split_attribute)]
# unique_list = node.nexts #.copy()
# for unique in unique_list:
# if (unique.name == val):
# feature_values_ = [element for i, element in enumerate(feature_values) if i != feature_names.index(split_attribute)]
# feature_names_ = [element for i, element in enumerate(feature_names) if element != split_attribute]
# return predict_from_node(feature_names_, feature_values_, unique.ptr)
# return None
if (not check_ifreal(X.iloc[:,0])):
def predict_from_node(X: pd.DataFrame, node):
if (node.decision is not None):
return node.decision
else:
split_attribute = node.to_split
split_attribute_val = X[split_attribute]
flag = 0
for i in node.nexts:
if i.name == split_attribute_val:
flag = 1
return predict_from_node(X,i.ptr)
if not flag:
return node.split_val
def predict_(X: pd.DataFrame, Tree_Root):
y = []
for i in range(X.shape[0]):
y.append(predict_from_node(X.iloc[i,:], Tree_Root))
return pd.Series(y)
# Traverse the tree you constructed to return the predicted values for the given test inputs.
else:
# for real inputs
def predict_from_node(dataFrame, root):
## Do check out for len(dataFrame.columns) == 0 condition
if (root.decision is not None):
return root.decision
split_attribute = root.to_split
split_value = root.split_val
if dataFrame[split_attribute][0] <= split_value:
return predict_from_node(dataFrame, root.left_ptr)
else:
return predict_from_node(dataFrame, root.right_ptr)
def predict_(X: pd.DataFrame, tree_root):
y = []
for i in range(X.shape[0]):
df = pd.DataFrame(X.iloc[i, :]).T
df.reset_index(inplace=True, drop=True)
y.append(predict_from_node(df, tree_root))
return pd.Series(y)
return predict_(X, self.Root)
def plot(self) -> None:
"""
Function to plot the tree
Output Example:
?(X1 > 4)
Y: ?(X2 > 7)
Y: Class A
N: Class B
N: Class C
Where Y => Yes and N => No
"""
def plot_(node_, depth):
if (node_.left_ptr is not None or node_.right_ptr is not None and node_.nexts == []):
return
else:
if (node_.to_split is None):
print(node_.decision)
return None
else: # node_decision is None
print('?(',node_.to_split,')', sep='')
for element in node_.nexts:
print(" "*depth, element.name, ': ',end = '', sep = '')
plot_(element.ptr, depth+1)
def _plot_(node_, depth):
if (node_.left_ptr == None and node_.right_ptr == None and node_.nexts != [] ):
return
else:
if (node_.decision is not None):
print(" "*depth,node_.decision)
return None
else: # node_decision is None
print(" "*depth,'?(', node_.to_split, '<=', node_.split_val, ')', sep='')
_plot_(node_.left_ptr, depth+1)
print(" "*depth,'?(', node_.to_split, '>', node_.split_val, ')', sep='')
_plot_(node_.right_ptr, depth+1)
plot_(self.Root, 0)
_plot_(self.Root,0)
return None
Decision Tree Implementation
Model Implementation
#test for DIDO
np.random.seed(42)
N = 30
P = 5
X = pd.DataFrame({i: pd.Series(np.random.randint(P, size=N), dtype="category") for i in range(5)})
y = pd.Series(np.random.randint(P, size=N), dtype="category")
num = 21
for criteria in ["information_gain", "gini_index"]:
tree = DecisionTree(criterion=criteria) # Split based on Inf. Gain
tree.fit(X.iloc[:num,:], y[:num])
y_hat = tree.predict(X.iloc[num:,:])
display(pd.concat((pd.DataFrame({'y_hat':y_hat}),pd.DataFrame({'y':list(y[num:])},index = list(range(0,N-num)))),axis=1))
tree.plot()
print("Criteria :", criteria)
print("Accuracy: ", accuracy(y_hat, y[num:]))
for cls in y.unique():
print(f'Class: {cls}')
print("Precision: ", precision(y_hat, y[num:], cls))
print("Recall: ", recall(y_hat, y[num:], cls))
print()| y_hat | y | |
|---|---|---|
| 0 | 0 | 0 |
| 1 | 0 | 1 |
| 2 | 0 | 3 |
| 3 | 2 | 3 |
| 4 | 3 | 1 |
| 5 | 0 | 2 |
| 6 | 0 | 0 |
| 7 | 0 | 4 |
| 8 | 0 | 0 |
?(3)
2: ?(0)
3: ?(1)
2: 0
1: 3
3: 3
2: 0
4: 0
1: 0
0: 0
0: ?(1)
3: 0
2: 2
1: 2
0: 0
4: ?(0)
2: 3
1: 1
1: ?(1)
4: 4
0: 0
1: 0
3: ?(2)
0: 2
4: 4
3: 2
Criteria : information_gain
Accuracy: 0.3333333333333333
Class: 0
Precision: 0.42857142857142855
Recall: 1.0
Class: 2
Precision: 0.0
Recall: 0.0
Class: 3
Precision: 0.0
Recall: 0.0
Class: 4
No predictions were made for class 4. Hence, precision is not defined.
Precision: None
Recall: 0.0
Class: 1
No predictions were made for class 1. Hence, precision is not defined.
Precision: None
Recall: 0.0
?(3)
2: ?(0)
3: ?(1)
2: 0
1: 3
3: 3
2: 0
4: 0
1: 0
0: 0
0: ?(1)
3: 0
2: 2
1: 2
0: 0
4: ?(0)
2: 3
1: 1
1: ?(1)
4: 4
0: 0
1: 0
3: ?(2)
0: 2
4: 4
3: 2
Criteria : gini_index
Accuracy: 0.3333333333333333
Class: 0
Precision: 0.42857142857142855
Recall: 1.0
Class: 2
Precision: 0.0
Recall: 0.0
Class: 3
Precision: 0.0
Recall: 0.0
Class: 4
No predictions were made for class 4. Hence, precision is not defined.
Precision: None
Recall: 0.0
Class: 1
No predictions were made for class 1. Hence, precision is not defined.
Precision: None
Recall: 0.0
| y_hat | y | |
|---|---|---|
| 0 | 0 | 0 |
| 1 | 0 | 1 |
| 2 | 0 | 3 |
| 3 | 2 | 3 |
| 4 | 3 | 1 |
| 5 | 0 | 2 |
| 6 | 0 | 0 |
| 7 | 0 | 4 |
| 8 | 0 | 0 |
# test for DIRO
np.random.seed(42)
N = 30
P = 5
X = pd.DataFrame({i: pd.Series(np.random.randint(P, size=N), dtype="category") for i in range(5)})
y = pd.Series(np.random.randn(N))
num = 21
for criteria in ["information_gain", "gini_index"]:
tree = DecisionTree(criterion=criteria) # Split based on Inf. Gain
tree.fit(X.iloc[:num,:], y[:num])
y_hat = tree.predict(X.iloc[num:,:])
display(pd.concat((pd.DataFrame({'y_hat':y_hat}),pd.DataFrame({'y':list(y[num:])},index = list(range(0,N-num)))),axis=1))
tree.plot()
print("Criteria :", criteria)
print("RMSE: ", rmse(y_hat, y[num:]))
print("MAE: ", mae(y_hat, y[num:]))| y_hat | y | |
|---|---|---|
| 0 | -0.146786 | -0.084165 |
| 1 | 0.301323 | 1.237016 |
| 2 | -0.146786 | -1.426479 |
| 3 | -0.146786 | 0.338023 |
| 4 | -0.146473 | 1.974571 |
| 5 | -0.146786 | 1.965035 |
| 6 | -0.846017 | -1.939911 |
| 7 | -0.146786 | -0.762627 |
| 8 | -0.146786 | 0.168659 |
?(1)
2: ?(0)
3: 0.724083251525877
4: -0.8703049545610739
2: -1.3099506872124742
3: ?(4)
3: ?(2)
3: -0.25576463695894236
4: 0.3013228009496343
4: ?(0)
2: 0.8499212040623495
3: 1.004986880702632
2: -0.6660606202304288
0: -0.4417660232260129
0: ?(3)
2: -1.3113242257511093
1: -1.0962657952813555
0: -0.5137874915082461
4: -0.14647280674161622
4: ?(0)
1: -0.5066432196315238
2: 2.943663416481284
4: -0.8224893687800624
1: ?(2)
3: ?(0)
4: 0.9148843219260121
0: 1.0867270230392738
1: -1.4585169209867577
4: -0.6627383133782134
0: -0.8460170164557059
Criteria : information_gain
RMSE: 1.2195665143747434
MAE: 1.0023179706204957
?(1)
2: ?(0)
3: 0.724083251525877
4: -0.8703049545610739
2: -1.3099506872124742
3: ?(0)
4: ?(2)
3: -0.25576463695894236
4: 0.3013228009496343
2: 0.8499212040623495
3: ?(2)
0: -0.6660606202304288
3: 1.004986880702632
1: -0.4417660232260129
0: ?(3)
2: -1.3113242257511093
1: -1.0962657952813555
0: -0.5137874915082461
4: -0.14647280674161622
4: ?(0)
1: -0.5066432196315238
2: 2.943663416481284
4: -0.8224893687800624
1: ?(0)
4: 0.9148843219260121
1: -1.4585169209867577
3: ?(2)
4: -0.6627383133782134
0: -0.8460170164557059
0: 1.0867270230392738
Criteria : gini_index
RMSE: 1.2013647062227255
MAE: 1.001281850071568| y_hat | y | |
|---|---|---|
| 0 | -0.146786 | -0.084165 |
| 1 | -0.146786 | 1.237016 |
| 2 | -0.146786 | -1.426479 |
| 3 | -0.146786 | 0.338023 |
| 4 | -0.146473 | 1.974571 |
| 5 | 0.849921 | 1.965035 |
| 6 | -1.458517 | -1.939911 |
| 7 | 1.004987 | -0.762627 |
| 8 | -0.146786 | 0.168659 |
# test for RIRO
np.random.seed(42)
N = 30
P = 5
X = pd.DataFrame(np.random.randn(N, P))
y = pd.Series(np.random.randn(N))
num = 21
for criteria in ["information_gain", "gini_index"]:
tree = DecisionTree(criterion=criteria) # Split based on Inf. Gain
tree.fit(X.iloc[:num,:], y[:num])
y_hat = tree.predict(X.iloc[num:,:])
display(pd.concat((pd.DataFrame({'y_hat':y_hat}),pd.DataFrame({'y':list(y[num:])},index = list(range(0,N-num)))),axis=1))
tree.plot()
print("Criteria :", criteria)
print("RMSE: ", rmse(y_hat, y[num:]))
print("MAE: ", mae(y_hat, y[num:]))| y_hat | y | |
|---|---|---|
| 0 | -0.245388 | -0.815810 |
| 1 | 0.250493 | -0.077102 |
| 2 | 1.158596 | 0.341152 |
| 3 | 0.293072 | 0.276691 |
| 4 | 0.721819 | 0.827183 |
| 5 | 0.473833 | 0.013002 |
| 6 | -0.897682 | 1.453534 |
| 7 | -0.245388 | -0.264657 |
| 8 | 1.158596 | 2.720169 |
?(1<=0.3506423035433175)
?(2.0<=-0.7919860527670518)
?(0.0<=-0.8500089439081561)
0.4738329209117875
?(0.0>-0.8500089439081561)
1.158595579007404
?(2.0>-0.7919860527670518)
?(3.0<=0.02125729698958631)
?(0.0<=0.78828567931571)
?(1.0<=-0.7392804369523404)
-0.24104886043268214
?(1.0>-0.7392804369523404)
-0.8786447031910382
?(0.0>0.78828567931571)
0.29307247329868125
?(3.0>0.02125729698958631)
?(3.0<=0.34644972822559905)
?(1.0<=-0.37608522749223366)
0.7218191061881409
?(1.0>-0.37608522749223366)
0.4127809269364983
?(3.0>0.34644972822559905)
?(3.0<=1.0673530726244467)
-0.897681994289516
?(3.0>1.0673530726244467)
0.25049285034587654
?(1>0.3506423035433175)
?(0.0<=-0.41792178458928875)
?(0.0<=-0.5517318279069667)
1.8657745111447566
?(0.0>-0.5517318279069667)
1.8967929826539474
?(0.0>-0.41792178458928875)
?(4.0<=-0.6003008780126726)
?(0.0<=0.070861858835451)
0.82206015999449
?(0.0>0.070861858835451)
0.9633761292443218
?(4.0>-0.6003008780126726)
?(0.0<=-0.06852970380057008)
0.3464482094969757
?(0.0>-0.06852970380057008)
-0.2453881160028705
Criteria : information_gain
RMSE: 1.0160790246579872
MAE: 0.6922328394475409
?(1<=0.3506423035433175)
?(2.0<=-0.7919860527670518)
?(0.0<=-0.8500089439081561)
0.4738329209117875
?(0.0>-0.8500089439081561)
1.158595579007404
?(2.0>-0.7919860527670518)
?(3.0<=0.02125729698958631)
?(0.0<=0.78828567931571)
?(1.0<=-0.7392804369523404)
-0.24104886043268214
?(1.0>-0.7392804369523404)
-0.8786447031910382
?(0.0>0.78828567931571)
0.29307247329868125
?(3.0>0.02125729698958631)
?(3.0<=0.34644972822559905)
?(1.0<=-0.37608522749223366)
0.7218191061881409
?(1.0>-0.37608522749223366)
0.4127809269364983
?(3.0>0.34644972822559905)
?(3.0<=1.0673530726244467)
-0.897681994289516
?(3.0>1.0673530726244467)
0.25049285034587654
?(1>0.3506423035433175)
?(0.0<=-0.41792178458928875)
?(0.0<=-0.5517318279069667)
1.8657745111447566
?(0.0>-0.5517318279069667)
1.8967929826539474
?(0.0>-0.41792178458928875)
?(4.0<=-0.6003008780126726)
?(0.0<=0.070861858835451)
0.82206015999449
?(0.0>0.070861858835451)
0.9633761292443218
?(4.0>-0.6003008780126726)
?(0.0<=-0.06852970380057008)
0.3464482094969757
?(0.0>-0.06852970380057008)
-0.2453881160028705
Criteria : gini_index
RMSE: 1.0160790246579872
MAE: 0.6922328394475409| y_hat | y | |
|---|---|---|
| 0 | -0.245388 | -0.815810 |
| 1 | 0.250493 | -0.077102 |
| 2 | 1.158596 | 0.341152 |
| 3 | 0.293072 | 0.276691 |
| 4 | 0.721819 | 0.827183 |
| 5 | 0.473833 | 0.013002 |
| 6 | -0.897682 | 1.453534 |
| 7 | -0.245388 | -0.264657 |
| 8 | 1.158596 | 2.720169 |
# test for RIDO
np.random.seed(42)
N = 30
P = 5
X = pd.DataFrame(np.random.randn(N, P))
y = pd.Series(np.random.randint(P, size=N), dtype="category")
num = 21
for criteria in ["information_gain", "gini_index"]:
tree = DecisionTree(criterion=criteria) # Split based on Inf. Gain
tree.fit(X.iloc[:num,:], y[:num])
y_hat = tree.predict(X.iloc[num:,:])
display(pd.concat((pd.DataFrame({'y_hat':y_hat}),pd.DataFrame({'y':list(y[num:])},index = list(range(0,N-num)))),axis=1))
tree.plot()
print("Criteria :", criteria)
print("Accuracy: ", accuracy(y_hat, y[num:]))
for cls in y.unique():
print(f'Class: {cls}')
print("Precision: ", precision(y_hat, y[num:], cls))
print("Recall: ", recall(y_hat, y[num:], cls))
print()| y_hat | y | |
|---|---|---|
| 0 | 2 | 4 |
| 1 | 3 | 3 |
| 2 | 4 | 2 |
| 3 | 3 | 0 |
| 4 | 0 | 0 |
| 5 | 0 | 3 |
| 6 | 1 | 2 |
| 7 | 0 | 2 |
| 8 | 0 | 4 |
?(4<=-1.4454128458513593)
?(0.0<=-0.5916309506035855)
0
?(0.0>-0.5916309506035855)
?(0.0<=0.7801845421853171)
2
?(0.0>0.7801845421853171)
4
?(4>-1.4454128458513593)
?(1.0<=1.7157455000081647)
?(4.0<=0.6775429329900813)
?(4.0<=-0.5947512395651535)
?(0.0<=0.5682840703666555)
0
?(0.0>0.5682840703666555)
1
?(4.0>-0.5947512395651535)
?(2.0<=-0.13759315591942284)
0
?(2.0>-0.13759315591942284)
3
?(4.0>0.6775429329900813)
?(3.0<=0.4714698601222159)
1
?(3.0>0.4714698601222159)
4
?(1.0>1.7157455000081647)
2
Criteria : information_gain
Accuracy: 0.2222222222222222
Class: 0
Precision: 0.25
Recall: 0.5
Class: 3
Precision: 0.5
Recall: 0.5
Class: 2
Precision: 0.0
Recall: 0.0
Class: 4
Precision: 0.0
Recall: 0.0
Class: 1
Precision: 0.0
No instance of class 1 was sampled. Hence, recall for class 1 is not defined.
Recall: None
?(4<=-1.4454128458513593)
?(0.0<=-0.5916309506035855)
0
?(0.0>-0.5916309506035855)
?(0.0<=0.7801845421853171)
2
?(0.0>0.7801845421853171)
4
?(4>-1.4454128458513593)
?(1.0<=1.7157455000081647)
?(4.0<=0.6775429329900813)
?(4.0<=-0.5947512395651535)
?(0.0<=0.5682840703666555)
0
?(0.0>0.5682840703666555)
1
?(4.0>-0.5947512395651535)
?(2.0<=-0.13759315591942284)
0
?(2.0>-0.13759315591942284)
3
?(4.0>0.6775429329900813)
?(3.0<=0.4714698601222159)
1
?(3.0>0.4714698601222159)
4
?(1.0>1.7157455000081647)
2
Criteria : gini_index
Accuracy: 0.2222222222222222
Class: 0
Precision: 0.25
Recall: 0.5
Class: 3
Precision: 0.5
Recall: 0.5
Class: 2
Precision: 0.0
Recall: 0.0
Class: 4
Precision: 0.0
Recall: 0.0
Class: 1
Precision: 0.0
No instance of class 1 was sampled. Hence, recall for class 1 is not defined.
Recall: None
| y_hat | y | |
|---|---|---|
| 0 | 2 | 4 |
| 1 | 3 | 3 |
| 2 | 4 | 2 |
| 3 | 3 | 0 |
| 4 | 0 | 0 |
| 5 | 0 | 3 |
| 6 | 1 | 2 |
| 7 | 0 | 2 |
| 8 | 0 | 4 |
from sklearn.datasets import make_blobs
X, y = make_blobs(n_samples=300, centers=4,
random_state=0, cluster_std=1.2)
plt.scatter(X[:, 0], X[:, 1], c=y, s=10, cmap='rainbow')
plt.xlabel(r"$x_1$")
plt.ylabel(r"$x_2$")Text(0, 0.5, '$x_2$')
# test for RIDO
np.random.seed(42)
N = 300
P = 4
X = pd.DataFrame(X)
y = pd.Series(y, dtype="category")
num = int(N*0.7)
for criteria in ["information_gain", "gini_index"]:
tree = DecisionTree(criterion=criteria) # Split based on Inf. Gain
tree.fit(X.iloc[:num,:], y[:num])
y_hat = tree.predict(X.iloc[num:,:])
display(pd.concat((pd.DataFrame({'y_hat':y_hat}),pd.DataFrame({'y':list(y[num:])},index = list(range(0,N-num)))),axis=1))
tree.plot()
print("Criteria :", criteria)
print("Accuracy: ", accuracy(y_hat, y[num:]))
for cls in y.unique():
print(f'Class: {cls}')
print("Precision: ", precision(y_hat, y[num:], cls))
print("Recall: ", recall(y_hat, y[num:], cls))
print()| y_hat | y | |
|---|---|---|
| 0 | 0 | 0 |
| 1 | 2 | 2 |
| 2 | 0 | 0 |
| 3 | 0 | 0 |
| 4 | 3 | 3 |
| ... | ... | ... |
| 85 | 0 | 1 |
| 86 | 1 | 1 |
| 87 | 3 | 0 |
| 88 | 3 | 3 |
| 89 | 2 | 2 |
90 rows × 2 columns
?(1<=3.744584695617445)
?(0.0<=-0.38840127416989967)
2
?(0.0>-0.38840127416989967)
?(1.0<=2.177059559323637)
1
?(1.0>2.177059559323637)
?(1.0<=3.6742256596476213)
?(0.0<=0.469146602388832)
0
?(0.0>0.469146602388832)
0
?(1.0>3.6742256596476213)
2
?(1>3.744584695617445)
?(1.0<=6.692241377858988)
?(0.0<=-0.9872897383437698)
?(1.0<=5.661767523442929)
?(1.0<=4.114941740589099)
2
?(1.0>4.114941740589099)
2
?(1.0>5.661767523442929)
?(1.0<=6.589666936280905)
3
?(1.0>6.589666936280905)
0
?(0.0>-0.9872897383437698)
?(1.0<=4.761010383499839)
?(1.0<=4.205715190188698)
0
?(1.0>4.205715190188698)
0
?(1.0>4.761010383499839)
?(1.0<=4.775306489805355)
3
?(1.0>4.775306489805355)
0
?(1.0>6.692241377858988)
3
Criteria : information_gain
Accuracy: 0.8777777777777778
Class: 1
Precision: 0.9230769230769231
Recall: 0.8888888888888888
Class: 3
Precision: 0.9523809523809523
Recall: 0.9090909090909091
Class: 0
Precision: 0.7037037037037037
Recall: 0.9047619047619048
Class: 2
Precision: 1.0
Recall: 0.8
?(1<=3.744584695617445)
?(0.0<=-0.38840127416989967)
2
?(0.0>-0.38840127416989967)
?(1.0<=2.177059559323637)
1
?(1.0>2.177059559323637)
?(1.0<=3.6742256596476213)
?(0.0<=0.469146602388832)
0
?(0.0>0.469146602388832)
0
?(1.0>3.6742256596476213)
2
?(1>3.744584695617445)
?(1.0<=6.692241377858988)
?(0.0<=-0.9872897383437698)
?(1.0<=5.661767523442929)
?(1.0<=4.114941740589099)
2
?(1.0>4.114941740589099)
2
?(1.0>5.661767523442929)
?(1.0<=6.589666936280905)
3
?(1.0>6.589666936280905)
0
?(0.0>-0.9872897383437698)
?(1.0<=4.761010383499839)
?(1.0<=4.205715190188698)
0
?(1.0>4.205715190188698)
0
?(1.0>4.761010383499839)
?(1.0<=4.775306489805355)
3
?(1.0>4.775306489805355)
0
?(1.0>6.692241377858988)
3
Criteria : gini_index
Accuracy: 0.8777777777777778
Class: 1
Precision: 0.9230769230769231
Recall: 0.8888888888888888
Class: 3
Precision: 0.9523809523809523
Recall: 0.9090909090909091
Class: 0
Precision: 0.7037037037037037
Recall: 0.9047619047619048
Class: 2
Precision: 1.0
Recall: 0.8
| y_hat | y | |
|---|---|---|
| 0 | 0 | 0 |
| 1 | 2 | 2 |
| 2 | 0 | 0 |
| 3 | 0 | 0 |
| 4 | 3 | 3 |
| ... | ... | ... |
| 85 | 0 | 1 |
| 86 | 1 | 1 |
| 87 | 3 | 0 |
| 88 | 3 | 3 |
| 89 | 2 | 2 |
90 rows × 2 columns
x = np.array([1, 2, 3, 4, 5, 6])
y = np.array([0, 0, 1, 1, 2, 2])
# test for RIRO
np.random.seed(42)
N = 6
X = pd.DataFrame(x)
y = pd.Series(y)
for criteria in ["information_gain", "gini_index"]:
tree = DecisionTree(criterion=criteria) # Split based on Inf. Gain
tree.fit(X, y)
y_hat = tree.predict(X)
display(pd.concat((pd.DataFrame({'y_hat':y_hat}),pd.DataFrame({'y':list(y)},index = list(range(0,N)))),axis=1))
tree.plot()
print("Criteria :", criteria)
print("RMSE: ", rmse(y_hat, y))
print("MAE: ", mae(y_hat, y))| y_hat | y | |
|---|---|---|
| 0 | 0 | 0 |
| 1 | 0 | 0 |
| 2 | 1 | 1 |
| 3 | 1 | 1 |
| 4 | 2 | 2 |
| 5 | 2 | 2 |
?(0<=2.5)
0
?(0>2.5)
?(0.0<=4.5)
1
?(0.0>4.5)
2
Criteria : information_gain
RMSE: 0.0
MAE: 0.0
?(0<=2.5)
0
?(0>2.5)
?(0.0<=4.5)
1
?(0.0>4.5)
2
Criteria : gini_index
RMSE: 0.0
MAE: 0.0| y_hat | y | |
|---|---|---|
| 0 | 0 | 0 |
| 1 | 0 | 0 |
| 2 | 1 | 1 |
| 3 | 1 | 1 |
| 4 | 2 | 2 |
| 5 | 2 | 2 |
t = pd.DataFrame(np.linspace(0,7,100))
t_pred = tree.predict(t)
plt.plot(t,t_pred)
plt.scatter(x,y)
plt.show()
x = np.linspace(0,7,100)
y = np.sin(x)
# test for RIRO
np.random.seed(42)
N = 100
X = pd.DataFrame(x)
y = pd.Series(y)
for criteria in ["information_gain", "gini_index"]:
tree = DecisionTree(criterion=criteria) # Split based on Inf. Gain
tree.fit(X, y)
y_hat = tree.predict(X)
display(pd.concat((pd.DataFrame({'y_hat':y_hat}),pd.DataFrame({'y':list(y)},index = list(range(0,N)))),axis=1))
tree.plot()
print("Criteria :", criteria)
print("RMSE: ", rmse(y_hat, y))
print("MAE: ", mae(y_hat, y))| y_hat | y | |
|---|---|---|
| 0 | 0.000000 | 0.000000 |
| 1 | 0.070648 | 0.070648 |
| 2 | 0.140943 | 0.140943 |
| 3 | 0.210534 | 0.210534 |
| 4 | 0.279073 | 0.279073 |
| ... | ... | ... |
| 95 | 0.387916 | 0.420491 |
| 96 | 0.513855 | 0.483539 |
| 97 | 0.513855 | 0.544171 |
| 98 | 0.629535 | 0.602083 |
| 99 | 0.629535 | 0.656987 |
100 rows × 2 columns
?(0<=3.005050505050505)
?(0.0<=0.5303030303030303)
?(0.0<=0.24747474747474746)
?(0.0<=0.10606060606060605)
?(0.0<=0.03535353535353535)
0.0
?(0.0>0.03535353535353535)
0.07064816888592165
?(0.0>0.10606060606060605)
?(0.0<=0.17676767676767674)
0.1409432800978515
?(0.0>0.17676767676767674)
0.21053404033479042
?(0.0>0.24747474747474746)
?(0.0<=0.38888888888888884)
?(0.0<=0.3181818181818182)
0.27907267622443427
?(0.0>0.3181818181818182)
0.34621667228835856
?(0.0>0.38888888888888884)
?(0.0<=0.45959595959595956)
0.41163048263136653
?(0.0>0.45959595959595956)
0.47498720780099063
?(0.0>0.5303030303030303)
?(0.0<=2.51010101010101)
?(0.0<=0.8838383838383838)
?(0.0<=0.6717171717171717)
0.565122507992744
?(0.0>0.6717171717171717)
0.7005289861202653
?(0.0>0.8838383838383838)
?(0.0<=2.1565656565656566)
0.9328597088095921
?(0.0>2.1565656565656566)
0.7194759387647931
?(0.0>2.51010101010101)
?(0.0<=2.792929292929293)
?(0.0<=2.6515151515151514)
0.5315184294679454
?(0.0>2.6515151515151514)
0.4069310816515149
?(0.0>2.792929292929293)
?(0.0<=2.9343434343434343)
0.2742194951535717
?(0.0>2.9343434343434343)
0.1710504011273846
?(0>3.005050505050505)
?(0.0<=6.116161616161616)
?(0.0<=3.641414141414141)
?(0.0<=3.287878787878788)
?(0.0<=3.146464646464646)
0.06574642171184002
?(0.0>3.146464646464646)
-0.07545994817996028
?(0.0>3.287878787878788)
?(0.0<=3.429292929292929)
-0.21515978623491833
?(0.0>3.429292929292929)
-0.3830250830635027
?(0.0>3.641414141414141)
?(0.0<=5.621212121212121)
?(0.0<=3.994949494949495)
-0.6230112638168462
?(0.0>3.994949494949495)
-0.8894871549832581
?(0.0>5.621212121212121)
?(0.0<=5.904040404040404)
-0.49581242890022725
?(0.0>5.904040404040404)
-0.2692534475993057
?(0.0>6.116161616161616)
?(0.0<=6.54040404040404)
?(0.0<=6.328282828282828)
?(0.0<=6.257575757575758)
-0.09610767710360636
?(0.0>6.257575757575758)
0.009743831559563916
?(0.0>6.328282828282828)
?(0.0<=6.398989898989899)
0.08036429967028173
?(0.0>6.398989898989899)
0.18531631945405186
?(0.0>6.54040404040404)
?(0.0<=6.752525252525253)
?(0.0<=6.611111111111111)
0.28841613698740426
?(0.0>6.611111111111111)
0.3879162237229862
?(0.0>6.752525252525253)
?(0.0<=6.893939393939394)
0.5138550232185234
?(0.0>6.893939393939394)
0.6295348844206806
Criteria : information_gain
RMSE: 0.06541273324008903
MAE: 0.04922901930948384
?(0<=3.005050505050505)
?(0.0<=0.5303030303030303)
?(0.0<=0.24747474747474746)
?(0.0<=0.10606060606060605)
?(0.0<=0.03535353535353535)
0.0
?(0.0>0.03535353535353535)
0.07064816888592165
?(0.0>0.10606060606060605)
?(0.0<=0.17676767676767674)
0.1409432800978515
?(0.0>0.17676767676767674)
0.21053404033479042
?(0.0>0.24747474747474746)
?(0.0<=0.38888888888888884)
?(0.0<=0.3181818181818182)
0.27907267622443427
?(0.0>0.3181818181818182)
0.34621667228835856
?(0.0>0.38888888888888884)
?(0.0<=0.45959595959595956)
0.41163048263136653
?(0.0>0.45959595959595956)
0.47498720780099063
?(0.0>0.5303030303030303)
?(0.0<=2.51010101010101)
?(0.0<=0.8838383838383838)
?(0.0<=0.6717171717171717)
0.565122507992744
?(0.0>0.6717171717171717)
0.7005289861202653
?(0.0>0.8838383838383838)
?(0.0<=2.1565656565656566)
0.9328597088095921
?(0.0>2.1565656565656566)
0.7194759387647931
?(0.0>2.51010101010101)
?(0.0<=2.792929292929293)
?(0.0<=2.6515151515151514)
0.5315184294679454
?(0.0>2.6515151515151514)
0.4069310816515149
?(0.0>2.792929292929293)
?(0.0<=2.9343434343434343)
0.2742194951535717
?(0.0>2.9343434343434343)
0.1710504011273846
?(0>3.005050505050505)
?(0.0<=6.116161616161616)
?(0.0<=3.641414141414141)
?(0.0<=3.287878787878788)
?(0.0<=3.146464646464646)
0.06574642171184002
?(0.0>3.146464646464646)
-0.07545994817996028
?(0.0>3.287878787878788)
?(0.0<=3.429292929292929)
-0.21515978623491833
?(0.0>3.429292929292929)
-0.3830250830635027
?(0.0>3.641414141414141)
?(0.0<=5.621212121212121)
?(0.0<=3.994949494949495)
-0.6230112638168462
?(0.0>3.994949494949495)
-0.8894871549832581
?(0.0>5.621212121212121)
?(0.0<=5.904040404040404)
-0.49581242890022725
?(0.0>5.904040404040404)
-0.2692534475993057
?(0.0>6.116161616161616)
?(0.0<=6.54040404040404)
?(0.0<=6.328282828282828)
?(0.0<=6.257575757575758)
-0.09610767710360636
?(0.0>6.257575757575758)
0.009743831559563916
?(0.0>6.328282828282828)
?(0.0<=6.398989898989899)
0.08036429967028173
?(0.0>6.398989898989899)
0.18531631945405186
?(0.0>6.54040404040404)
?(0.0<=6.752525252525253)
?(0.0<=6.611111111111111)
0.28841613698740426
?(0.0>6.611111111111111)
0.3879162237229862
?(0.0>6.752525252525253)
?(0.0<=6.893939393939394)
0.5138550232185234
?(0.0>6.893939393939394)
0.6295348844206806
Criteria : gini_index
RMSE: 0.06541273324008903
MAE: 0.04922901930948384| y_hat | y | |
|---|---|---|
| 0 | 0.000000 | 0.000000 |
| 1 | 0.070648 | 0.070648 |
| 2 | 0.140943 | 0.140943 |
| 3 | 0.210534 | 0.210534 |
| 4 | 0.279073 | 0.279073 |
| ... | ... | ... |
| 95 | 0.387916 | 0.420491 |
| 96 | 0.513855 | 0.483539 |
| 97 | 0.513855 | 0.544171 |
| 98 | 0.629535 | 0.602083 |
| 99 | 0.629535 | 0.656987 |
100 rows × 2 columns
t = pd.DataFrame(np.linspace(0,7,100))
t_pred = tree.predict(t)
plt.plot(t,t_pred)
plt.plot(x,y)
plt.show()