Decision Trees Tests

This is an example of using my implementation of a Decision Tree. My implementation is based of of this example and modified to use pandas for data processing.

This example looks at bank note data to classify if a note is real or fake. The data is sensor data and the target is the class we are trying to predict.

import pandas as pd
pd.read_csv("datasets/data_banknote_authentication.csv", names=['f1', 'f2', 'f3', 'f4', 'target'])

	f1	f2	f3	f4	target
0	3.62160	8.66610	-2.8073	-0.44699	0
1	4.54590	8.16740	-2.4586	-1.46210	0
2	3.86600	-2.63830	1.9242	0.10645	0
3	3.45660	9.52280	-4.0112	-3.59440	0
4	0.32924	-4.45520	4.5718	-0.98880	0
...	...	...	...	...	...
1367	0.40614	1.34920	-1.4501	-0.55949	1
1368	-1.38870	-4.87730	6.4774	0.34179	1
1369	-3.75030	-13.45860	17.5932	-2.77710	1
1370	-3.56370	-8.38270	12.3930	-1.28230	1
1371	-2.54190	-0.65804	2.6842	1.19520	1

1372 rows × 5 columns

import random
from decision_tree import DecisionTree

random.seed(1)
df = pd.read_csv("./datasets/data_banknote_authentication.csv", header=None)
df = df.sample(frac=1, random_state=1).reset_index(drop=True) # data is shuffled
Xs = df.iloc[:, :-1]
ys = df.iloc[:, -1].to_frame()
Xs

	0	1	2	3
0	-3.55100	1.89550	0.186500	-2.440900
1	1.31140	4.54620	2.293500	0.225410
2	-4.01730	-8.31230	12.454700	-1.437500
3	-5.11900	6.64860	-0.049987	-6.520600
4	3.62890	0.81322	1.627700	0.776270
...	...	...	...	...
1367	3.49160	8.57090	-3.032600	-0.591820
1368	0.74521	3.63570	-4.404400	-4.141400
1369	-4.36670	6.06920	0.572080	-5.466800
1370	2.04660	2.03000	2.176100	-0.083634
1371	-2.31470	3.66680	-0.696900	-1.247400

1372 rows × 4 columns

ys

	4
0	1
1	0
2	1
3	1
4	0
...	...
1367	0
1368	1
1369	1
1370	0
1371	1

1372 rows × 1 columns

Next the data is split into a train and test set

splitInd = int(len(Xs) * 0.8)

trainX = Xs.iloc[0:splitInd]
testX  = Xs.iloc[splitInd:]
trainY = ys.iloc[0:splitInd]
testY  = ys.iloc[splitInd:]

trainX, testY

(           0        1          2         3
 0    -3.5510  1.89550   0.186500 -2.440900
 1     1.3114  4.54620   2.293500  0.225410
 2    -4.0173 -8.31230  12.454700 -1.437500
 3    -5.1190  6.64860  -0.049987 -6.520600
 4     3.6289  0.81322   1.627700  0.776270
 ...      ...      ...        ...       ...
 1092 -1.3887 -4.87730   6.477400  0.341790
 1093  1.5701  7.91290   0.290180 -2.195300
 1094  1.0135  8.45510  -1.672000 -2.081500
 1095  0.3798  0.70980   0.757200 -0.444400
 1096 -1.8219 -6.88240   5.468100  0.057313
 
 [1097 rows x 4 columns],
       4
 1097  0
 1098  1
 1099  0
 1100  0
 1101  0
 ...  ..
 1367  0
 1368  1
 1369  1
 1370  0
 1371  1
 
 [275 rows x 1 columns])

Next the tree will be created and fit to the test data

dt = DecisionTree()
dt.fit(trainX, trainY)
dt.printTree()

C:\Users\sam_m\Documents\machine-learning-trees\decision_tree.py:15: SettingWithCopyWarning: 
A value is trying to be set on a copy of a slice from a DataFrame.
Try using .loc[row_indexer,col_indexer] = value instead

See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy
  self.Xs['$target$'] = self.ys.iloc[:, 0]

[X0 < 0.322]
 [X1 < 7.627]
  [X0 < -0.398]
   [X2 < 6.220]
    [X0 < -3.551]
     [1.0]
     [1.0]
    [X1 < -4.606]
     [1.0]
     [0.0]
   [X1 < 5.897]
    [X2 < 3.114]
     [1.0]
     [0.0]
    [0.0]
  [X0 < -4.286]
   [X0 < -5.490]
    [1.0]
    [1.0]
   [X0 < -1.180]
    [X0 < -1.327]
     [0.0]
     [0.0]
    [X0 < -1.180]
     [0.0]
     [0.0]
 [X2 < -4.413]
  [X0 < 4.407]
   [X0 < 2.392]
    [X0 < 0.816]
     [1.0]
     [1.0]
    [1.0]
   [0.0]
  [X0 < 1.594]
   [X2 < -2.272]
    [X1 < 7.638]
     [1.0]
     [0.0]
    [X3 < 0.097]
     [0.0]
     [0.0]
   [X0 < 2.042]
    [X2 < -2.339]
     [1.0]
     [0.0]
    [X0 < 3.629]
     [0.0]
     [0.0]

Now a prediction will is performed and the accuracy is calculated

p = dt.predict(testX)

c = 0
testY.reset_index(drop=True, inplace=True)
for i, r in testY.iterrows():
    if p[i] == testY.iat[i, 0]:
        c += 1

print("Accuracy:", c / len(p))

Accuracy: 0.9709090909090909

On this data, the decision tree is about 97% accurate. This data works well for analysis with a decision tree.

This project was a great way for me to learn about decision trees and to expand my knowledge of machine learning.