DataBaser.Net: kNN

#iris 데이터세트 만들기
import numpy as np
import pandas as pd
from sklearn.datasets import load_iris
iris = load_iris()
iris.data
iris.feature_names
iris.target
iris.target_names


iris_df = pd.DataFrame(iris.data, columns=iris.feature_names)
iris_df["target"] = iris.target 
iris_df["target_names"] = iris.target_names[iris.target]
iris_df[:5]


#훈련세트, 테스트세트 나누기
from sklearn.model_selection import train_test_split
train_set, test_set = train_test_split(iris_df, test_size = 0.3)

train_set.shape
test_set.shape

#데이터 살펴보기
pd.tools.plotting.scatter_matrix(train_set, c=train_set.target, 
                                 figsize=(15,15), marker="o", 
                                 hist_kwds={"bins":20},s=60, alpha=0.8, cmap=mglearn.cm3)

#s: marker 크기
#cmap: color map


#kNN
import sklearn.neighbors as nn
knn = nn.KNeighborsClassifier(n_neighbors = 1)

#훈련
knn.fit(X=train_set.ix[:, [0,1,2,3]], y=train_set.target)


#테스트1
knn.score(test_set.ix[:, [0,1,2,3]], test_set.target) #정확도 95.6%


#테스트2
pred = knn.predict(X=test_set.ix[:, [0,1,2,3]])


# consusion matrix는 다음을 참고
# https://uberpython.wordpress.com/2012/01/01/precision-recall-sensitivity-and-specificity/
# https://stackoverflow.com/questions/31324218/scikit-learn-how-to-obtain-true-positive-true-negative-false-positive-and-fal
from pandas_ml import ConfusionMatrix
cm = ConfusionMatrix(test_set.target.values, pred)
cm.print_stats()

결과

cm.print_stats()
Confusion Matrix:

Predicted   0   1   2  __all__
Actual                        
0          16   0   0       16
1           0  16   0       16
2           0   2  11       13
__all__    16  18  11       45


Overall Statistics:

Accuracy: 0.955555555556
95% CI: (0.84850709975666083, 0.99457151129974908)
No Information Rate: ToDo
P-Value [Acc > NIR]: 2.8423103302e-15
Kappa: 0.932735426009
Mcnemar's Test P-Value: ToDo


Class Statistics:

Classes                                       0          1          2
Population                                   45         45         45
P: Condition positive                        16         16         13
N: Condition negative                        29         29         32
Test outcome positive                        16         18         11
Test outcome negative                        29         27         34
TP: True Positive                            16         16         11
TN: True Negative                            29         27         32
FP: False Positive                            0          2          0
FN: False Negative                            0          0          2
TPR: (Sensitivity, hit rate, recall)          1          1   0.846154
TNR=SPC: (Specificity)                        1   0.931034          1
PPV: Pos Pred Value (Precision)               1   0.888889          1
NPV: Neg Pred Value                           1          1   0.941176
FPR: False-out                                0  0.0689655          0
FDR: False Discovery Rate                     0   0.111111          0
FNR: Miss Rate                                0          0   0.153846
ACC: Accuracy                                 1   0.955556   0.955556
F1 score                                      1   0.941176   0.916667
MCC: Matthews correlation coefficient         1   0.909718   0.892401
Informedness                                  1   0.931034   0.846154
Markedness                                    1   0.888889   0.941176
Prevalence                             0.355556   0.355556   0.288889
LR+: Positive likelihood ratio              inf       14.5        inf
LR-: Negative likelihood ratio                0          0   0.153846
DOR: Diagnostic odds ratio                  inf        inf        inf
FOR: False omission rate                      0          0  0.0588235

CategoryMachineLearning