KNN으로 citrus data set 분류하기

 

 

02-knn-bin-citrus.ipynb

0.22MB

 

KNN (이진분류)

• citrus data set

 

In [1]:

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns

 

 

1. 데이터 준비¶

 

In [2]:

from google.colab import drive
drive.mount('/content/drive')

 

 

Mounted at /content/drive

 

In [97]:

df = pd.read_csv("/content/drive/MyDrive/SKT FLY AI/2주차/citrus.csv")

 

In [77]:

df.head()

 

Out[77]:

	name	diameter	weight	red	green	blue
0	orange	2.96	86.76	172	85	2
1	orange	3.91	88.05	166	78	3
2	orange	4.42	95.17	156	81	2
3	orange	4.47	95.60	163	81	4
4	orange	4.48	95.76	161	72	9

 

 

 

In [68]:

# 데이터의 shape
df.shape

 

Out[68]:

(10000, 6)

 

In [69]:

df.info()

 

 

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 10000 entries, 0 to 9999
Data columns (total 6 columns):
 #   Column    Non-Null Count  Dtype  
---  ------    --------------  -----  
 0   name      10000 non-null  object 
 1   diameter  10000 non-null  float64
 2   weight    10000 non-null  float64
 3   red       10000 non-null  int64  
 4   green     10000 non-null  int64  
 5   blue      10000 non-null  int64  
dtypes: float64(2), int64(3), object(1)
memory usage: 468.9+ KB

 

In [70]:

df.name.unique()

 

Out[70]:

array(['orange', 'grapefruit'], dtype=object)

 

In [71]:

df.describe()

 

Out[71]:

	diameter	weight	red	green	blue
count	10000.000000	10000.000000	10000.000000	10000.000000	10000.000000
mean	9.975685	175.050792	153.847800	76.010600	11.363200
std	1.947844	29.212119	10.432954	11.708433	9.061275
min	2.960000	86.760000	115.000000	31.000000	2.000000
25%	8.460000	152.220000	147.000000	68.000000	2.000000
50%	9.980000	174.985000	154.000000	76.000000	10.000000
75%	11.480000	197.722500	161.000000	84.000000	17.000000
max	16.450000	261.510000	192.000000	116.000000	56.000000

 

 

 

 

In [72]:

# 결측치(Missing value)

df.isna().sum(axis=0)

 

Out[72]:

name        0
diameter    0
weight      0
red         0
green       0
blue        0
dtype: int64

 

In [73]:

df.name.value_counts()

 

Out[73]:

orange        5000
grapefruit    5000
Name: name, dtype: int64

 

In [98]:

df = df.sample(frac =1)
df.head(10)

 

Out[98]:

	name	diameter	weight	red	green	blue
6188	grapefruit	10.62	183.46	156	58	13
1608	orange	7.90	144.19	159	80	2
7879	grapefruit	11.69	201.17	147	57	19
6552	grapefruit	10.86	187.98	149	53	18
6556	grapefruit	10.87	188.01	139	50	4
6039	grapefruit	10.48	181.57	146	67	21
5445	grapefruit	9.82	171.60	153	69	2
568	orange	6.95	130.84	151	92	13
2100	orange	8.23	148.97	158	90	29
547	orange	6.93	130.45	170	76	9

 

 

 

In [99]:

df.reset_index(drop=True, inplace=True)
df

 

Out[99]:

	name	diameter	weight	red	green	blue
0	grapefruit	10.62	183.46	156	58	13
1	orange	7.90	144.19	159	80	2
2	grapefruit	11.69	201.17	147	57	19
3	grapefruit	10.86	187.98	149	53	18
4	grapefruit	10.87	188.01	139	50	4
...	...	...	...	...	...	...
9995	grapefruit	13.06	221.37	157	61	7
9996	grapefruit	11.22	193.83	159	73	25
9997	grapefruit	12.56	214.02	168	54	19
9998	orange	9.48	168.20	149	76	19
9999	orange	8.57	153.77	151	104	2

10000 rows × 6 columns

 

 

 

 

In [120]:

# 테스트 데이터 분리

X = df.loc[:7999]
y = df.loc[8000:]

 

In [117]:

X

 

Out[117]:

	name	diameter	weight	red	green	blue
0	grapefruit	10.62	183.46	156	58	13
1	orange	7.90	144.19	159	80	2
2	grapefruit	11.69	201.17	147	57	19
3	grapefruit	10.86	187.98	149	53	18
4	grapefruit	10.87	188.01	139	50	4
...	...	...	...	...	...	...
7995	grapefruit	10.50	181.80	143	89	13
7996	orange	7.79	142.14	176	74	6
7997	orange	8.20	148.54	173	82	2
7998	orange	8.19	148.41	162	80	2
7999	grapefruit	11.81	202.95	155	72	4

8000 rows × 6 columns

 

In [121]:

y

 

Out[121]:

	name	diameter	weight	red	green	blue
8000	orange	7.87	143.73	145	87	16
8001	orange	8.37	150.84	155	83	6
8002	grapefruit	10.92	188.98	171	69	30
8003	orange	7.49	137.99	168	69	2
8004	grapefruit	11.30	195.19	148	80	32
...	...	...	...	...	...	...
9995	grapefruit	13.06	221.37	157	61	7
9996	grapefruit	11.22	193.83	159	73	25
9997	grapefruit	12.56	214.02	168	54	19
9998	orange	9.48	168.20	149	76	19
9999	orange	8.57	153.77	151	104	2

2000 rows × 6 columns

 

 

 

 

전처리¶

스케일링

 

In [147]:

X.columns

 

Out[147]:

Index(['name', 'diameter', 'weight', 'red', 'green', 'blue'], dtype='object')

 

In [148]:

X_target = X.name
X_train = X[['diameter', 'weight', 'red', 'green', 'blue']]

 

In [149]:

X_target

 

Out[149]:

0       grapefruit
1           orange
2       grapefruit
3       grapefruit
4       grapefruit
           ...    
7995    grapefruit
7996        orange
7997        orange
7998        orange
7999    grapefruit
Name: name, Length: 8000, dtype: object

 

In [150]:

X_train

 

Out[150]:

	diameter	weight	red	green	blue
0	10.62	183.46	156	58	13
1	7.90	144.19	159	80	2
2	11.69	201.17	147	57	19
3	10.86	187.98	149	53	18
4	10.87	188.01	139	50	4
...	...	...	...	...	...
7995	10.50	181.80	143	89	13
7996	7.79	142.14	176	74	6
7997	8.20	148.54	173	82	2
7998	8.19	148.41	162	80	2
7999	11.81	202.95	155	72	4

8000 rows × 5 columns

 

 

 

 

평균 및 분산을 통해 Z정규화¶

 

In [151]:

u_dia = X_train.diameter.mean()
std_dia = X_train.diameter.std()
X_train.diameter = (X_train.diameter - u_dia )/std_dia

 

 

<ipython-input-151-20b9ae9e183a>:3: 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
  X_train.diameter = (X_train.diameter - u_dia )/std_dia

 

In [152]:

u_wei = X_train.weight.mean()
std_wei = X_train.weight.std()
X_train.weight = (X_train.weight - u_wei )/std_wei

 

 

<ipython-input-152-7de1ec988aee>:3: 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
  X_train.weight = (X_train.weight - u_wei )/std_wei

 

In [153]:

u_r = X_train.red.mean()
std_r = X_train.red.std()
X_train.red = (X_train.red - u_r )/std_r

 

 

<ipython-input-153-af9162887f7b>:3: 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
  X_train.red = (X_train.red - u_r )/std_r

 

In [154]:

u_g = X_train.green.mean()
std_g = X_train.green.std()
X_train.green = (X_train.green - u_g )/std_g

 

 

<ipython-input-154-ddc5e8fc4fb8>:3: 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
  X_train.green = (X_train.green - u_g )/std_g

 

In [155]:

u_b = X_train.blue.mean()
std_b = X_train.blue.std()
X_train.blue = (X_train.blue - u_b )/std_b

 

 

<ipython-input-155-db2495ad9551>:3: 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
  X_train.blue = (X_train.blue - u_b )/std_b

 

In [156]:

X_train

 

Out[156]:

	diameter	weight	red	green	blue
0	0.333416	0.290576	0.205745	-1.544606	0.177122
1	-1.064225	-1.054454	0.491651	0.338238	-1.034816
2	0.883224	0.897159	-0.651973	-1.630190	0.838179
3	0.456738	0.445390	-0.461369	-1.972525	0.728003
4	0.461876	0.446418	-1.414389	-2.229277	-0.814464
...	...	...	...	...	...
7995	0.271756	0.233720	-1.033181	1.108492	0.177122
7996	-1.120747	-1.124668	2.111786	-0.175265	-0.594112
7997	-0.910073	-0.905462	1.825879	0.509406	-1.034816
7998	-0.915211	-0.909915	0.777557	0.338238	-1.034816
7999	0.944884	0.958125	0.110443	-0.346433	-0.814464

8000 rows × 5 columns

 

 

 

 

In [159]:

X_target = X_target.map({'grapefruit': 0, 'orange':1})
X_target

 

Out[159]:

0       0
1       1
2       0
3       0
4       0
       ..
7995    0
7996    1
7997    1
7998    1
7999    0
Name: name, Length: 8000, dtype: int64

 

 

학습¶

• K-nn 알고리즘

 

In [160]:

type(X_train),type(X_target)

 

Out[160]:

(pandas.core.frame.DataFrame, pandas.core.series.Series)

 

In [162]:

# Numpy array 형식으로 변환
X_train = X_train.values
X_target = X_target.values

 

In [164]:

# 베이스 모델 학습
from sklearn.neighbors import KNeighborsClassifier

clf = KNeighborsClassifier()
clf = clf.fit(X_train,X_target)

 

In [166]:

y_target = y.name
y_train = y[['diameter','weight','red','green','blue']]
y_train

 

Out[166]:

	diameter	weight	red	green	blue
8000	7.87	143.73	145	87	16
8001	8.37	150.84	155	83	6
8002	10.92	188.98	171	69	30
8003	7.49	137.99	168	69	2
8004	11.30	195.19	148	80	32
...	...	...	...	...	...
9995	13.06	221.37	157	61	7
9996	11.22	193.83	159	73	25
9997	12.56	214.02	168	54	19
9998	9.48	168.20	149	76	19
9999	8.57	153.77	151	104	2

2000 rows × 5 columns

 

 

 

 

In [167]:

y_target = y_target.map({'grapefruit': 0, 'orange':1})

 

In [168]:

y_train.diameter = (y_train.diameter - u_dia )/std_dia

 

 

 

 

In [169]:

y_train.weight = (y_train.weight - u_wei )/std_wei

 

 

 

 

In [170]:

y_train.red = (y_train.red - u_r )/std_r

 

 

 

 

In [171]:

y_train.green = (y_train.green - u_g )/std_g

 

 

 

 

In [172]:

y_train.blue = (y_train.blue - u_b )/std_b

 

 

 

 

 

 

In [175]:

y_train = y_train.values

 

In [176]:

y_target = y_target.values

 

In [177]:

# 예측 진행

y_pred = clf.predict(y_train)

 

In [180]:

y_pred[:20]

 

Out[180]:

array([1, 1, 0, 1, 0, 0, 1, 0, 1, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0])

 

In [181]:

y_target[:20]

 

Out[181]:

array([1, 1, 0, 1, 0, 0, 1, 0, 1, 1, 0, 0, 0, 1, 1, 1, 0, 0, 0, 0])

 

 

평가¶

 

In [182]:

# 정확도 평가

np.sum(y_pred == y_target) / len(y_pred)

 

Out[182]:

0.9205

 

In [185]:

from sklearn.metrics import confusion_matrix

cf = confusion_matrix(y_target, y_pred)
cf

 

Out[185]:

array([[929,  73],
       [ 86, 912]])

 

In [186]:

s = sns.heatmap(cf, annot = True,cmap='Blues')

 

 

 

In [192]:

from sklearn.metrics import accuracy_score, recall_score, precision_score,f1_score


acc = accuracy_score(y_target,y_pred) # 정답값, 예측값
recall = recall_score(y_target,y_pred) # 정답값, 예측값
precision = precision_score(y_target,y_pred) # 정답값, 예측값
f1 = f1_score(y_target,y_pred) # 정답값, 예측값

print(f'accuracy : {acc}')
print(f'recall : {round(recall,3)}')
print(f'precision : {round(precision,3)}')
print(f'f1 : {round(f1,3)}')

 

 

accuracy : 0.9205
recall : 0.914
precision : 0.926
f1 : 0.92

 

 

모델 튜닝¶

 

In [195]:

score = []

for k in range(3,100):
  clf = KNeighborsClassifier(n_neighbors = k)
  clf.fit(X_train,X_target)
  y_pred = clf.predict(y_train)
  acc = accuracy_score(y_target,y_pred) # 정답값, 예측값
  score.append(acc)
print(f'max_accuracy: {max(score)}')

 

 

max_accuracy: 0.9355

 

In [198]:

plt.plot(range(3,100),score)
plt.show()

 

 

 

In [199]:

# k가 특정 N(40)일 때가 베스트라 가정, 재학습을 진행

best_clf = KNeighborsClassifier(n_neighbors = 60)
best_clf.fit(X_train,X_target)
y_pred = best_clf.predict(y_train)
acc = accuracy_score(y_target,y_pred) # 정답값, 예측

print(f'optimal_accuracy: {acc}')

 

 

optimal_accuracy: 0.935