In [23]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
In [24]:
from sklearn import model_selection
from sklearn import preprocessing
from sklearn import metrics
from sklearn import tree
In [25]:
data = pd.read_csv('diabetes.csv')
In [26]:
y = data['Outcome']
X = data.drop(columns=['Outcome'], axis=1)

Trening, validacioni i test skup ćemo u ovom primeru podeliti u odnosu 56:14:30.

In [27]:
X_train_and_val, X_test, y_train_and_val, y_test = model_selection.train_test_split(
    X, y, test_size = 0.3, random_state = 7, stratify = y)
In [28]:
X_train, X_val, y_train, y_val = model_selection.train_test_split(
    X_train_and_val, y_train_and_val, train_size = 0.8,
    random_state = 7, stratify = y_train_and_val)
In [29]:
scaler = preprocessing.StandardScaler()
scaler.fit(X_train)
X_train = scaler.transform(X_train)
X_val = scaler.transform(X_val)
X_test = scaler.transform(X_test)
In [30]:
max_depths = [2,3,4,5,6,7]
In [31]:
criterions = ['gini', 'entropy', 'log_loss']
In [32]:
max_featuress = [0.75, 0.8, 0.85, 0.9, 0.95]
In [33]:
best_f1_score = 0
best_max_depth = None
best_criterion = None
best_max_feature = None
In [34]:
for max_depth in max_depths:
    for criterion in criterions:
        for max_features in max_featuress:
            model = tree.DecisionTreeClassifier(criterion=criterion,
                                                max_features=max_features,
                                                max_depth=max_depth, random_state=7)
            model.fit(X_train, y_train)
            f1_score = metrics.f1_score(y_val, model.predict(X_val))
            if f1_score > best_f1_score:
                best_f1_score = f1_score
                best_max_depth = max_depth
                best_criterion = criterion
                best_max_features = max_features
In [35]:
print('Najbolji f1-score na validacionom skupu je: ', best_f1_score)

Najbolji f1-score na validacionom skupu je:  0.5952380952380952
In [36]:
print('Najbolji hiperparametri modela su: ', best_max_depth,
      best_criterion, best_max_features)

Najbolji hiperparametri modela su:  2 gini 0.75

Kreiramo konačni model, koristeći trening skup:

In [37]:
best_model = tree.DecisionTreeClassifier(criterion=best_criterion,
                                                max_features=best_max_features,
                                                max_depth=best_max_depth,
                                                random_state=7)
best_model.fit(X_train, y_train)
Out[37]:
DecisionTreeClassifier(max_depth=2, max_features=0.75, random_state=7)
In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.
DecisionTreeClassifier(max_depth=2, max_features=0.75, random_state=7)
In [38]:
f1_score = metrics.f1_score(y_test, best_model.predict(X_test))
In [39]:
print('f1-score na test skupu je: ', f1_score)

f1-score na test skupu je:  0.6040268456375839