[AI] 와인 종류 예측

와인 종류 : red, white 두가지 타입으로 예측

 

winequality-white.csv

0.25MB

winequality-red.csv

0.08MB

import pandas as pd

# 데이터 로드
# 원본 파일은 분리자가 세미콜론
red_df = pd.read_csv('./winequality-red.csv', sep=';')
white_df = pd.read_csv('./winequality-white.csv', sep=';')

# 분리자를 콤마로 사본 저장 (일반적으로 csv 파일은 세미콜론이 아닌 콤마로 분리자를 사용)
red_df.to_csv('./winequality-red2.csv', index=False)
white_df.to_csv('./winequality-white2.csv', index=False)

red_df = pd.read_csv('./winequality-red2.csv')
white_df = pd.read_csv('./winequality-white2.csv')

# red wine 표준화 시행
from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()

red_scaled_np = scaler.fit_transform(red_df)

# scale 리턴값은 numpy 이므로 다시 DataFrame 생성
red_scaled_df = pd.DataFrame(red_scaled_np, columns=red_df.columns)

red_scaled_df.describe()

# white wine 표준화 시행
white_scaled_np = scaler.fit_transform(white_df)

# scale 리턴값은 numpy 이므로 다시 DataFrame 생성
white_scaled_df = pd.DataFrame(white_scaled_np, columns=white_df.columns)

white_scaled_df.describe()

# 구분을 위해 type 열 추가
print('before => ',red_scaled_df.shape, white_scaled_df.shape)

red_scaled_df['type'] = 'red'
white_scaled_df['type'] = 'white'

print('after => ',red_scaled_df.shape, white_scaled_df.shape)

white_scaled_df.head()
red_scaled_df.head()

# pd.concat() 이용해서 하나로 합치기
wine_scaled_df = pd.concat([red_scaled_df, white_scaled_df],axis=0)

wine_scaled_df.shape

wine_scaled_df['type']

wine_scaled_df.info()

wine_scaled_df.describe()

wine_scaled_df.isnull().sum()

wine_scaled_df['type'].unique()

wine_scaled_df['type'].value_counts()

wine_scaled_df['type'].isna()

# type 항목의 'red' = 0, white = 1로 변경
wine_scaled_df['type'] = wine_scaled_df['type'].replace('red', 0)
wine_scaled_df['type'] = wine_scaled_df['type'].replace('white', 1)

wine_scaled_df.head()

# 입력 데이터
feature_col = wine_scaled_df.columns.difference(['type'])

# 정답 데이터
label_col = wine_scaled_df['type']

wine_feature_df = wine_scaled_df[feature_col]
wine_label_df = label_col

wine_feature_df.head()

wine_label_df.head()

import numpy as np

wine_feature_np = wine_feature_df.to_numpy().astype('float32')
wine_label_np = wine_label_df.to_numpy().astype('float32')

print(wine_feature_np.shape, wine_label_np.shape)

# shuffle
s = np.arange(len(wine_feature_np))

np.random.shuffle(s)

wine_feature_np = wine_feature_np[s]
wine_label_np = wine_label_np[s]

print(wine_feature_np.shape, wine_label_np.shape)

# 85% : 15% 비율로 train / test data 분리
split_ratio = 0.15

test_num = int(split_ratio*len(wine_feature_np))

x_test = wine_feature_np[:test_num]
y_test = wine_label_np[:test_num]

x_train = wine_feature_np[test_num:]
y_train = wine_label_np[test_num:]

import tensorflow as tf
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Dropout
from tensorflow.keras.optimizers import Adam
import matplotlib.pyplot as plt

model = Sequential()

model.add(Dense(128, activation='relu', input_shape=(x_train.shape[1],)))

model.add(Dense(1, activation='sigmoid'))  # type

model.summary()

Model: "sequential"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 dense (Dense)               (None, 128)               1664      
                                                                 
 dense_1 (Dense)             (None, 1)                 129       
                                                                 
=================================================================
Total params: 1,793
Trainable params: 1,793
Non-trainable params: 0
_________________________________________________________________

model.compile(optimizer=tf.keras.optimizers.Adam(),
              loss='binary_crossentropy', metrics=['accuracy'] )

from datetime import datetime

start_time = datetime.now()

hist = model.fit(x_train, y_train, 
                 epochs=200, batch_size=32, 
                 validation_data=(x_test, y_test))

end_time = datetime.now()

print('elapsed time => ', end_time-start_time)

pred = model.predict(x_test[-5:])

print(pred.flatten())
print(y_test[-5:])

[0.4170394  0.99991196 1.         0.4412927  0.98263514]
[1. 1. 1. 0. 0.]

model.evaluate(x_test, y_test)

31/31 [==============================] - 0s 3ms/step - loss: 0.1381 - accuracy: 0.9630
[0.13809409737586975, 0.9630390405654907]

plt.plot(hist.history['loss'], label='train')
plt.plot(hist.history['val_loss'], label='validation')
plt.title('loss trend')
plt.ylabel('loss')
plt.xlabel('epoch')
plt.legend(loc='best')
plt.grid()
plt.show()

plt.plot(hist.history['accuracy'], label='train')
plt.plot(hist.history['val_accuracy'], label='validation')
plt.title('accuracy trend')
plt.ylabel('accuracy')
plt.xlabel('epoch')
plt.legend(loc='best')
plt.grid()
plt.show()