基於keras的深度學習——分類
阿新 • • 發佈：2018-12-31
使用keras的深度學習來分類白葡萄酒還是紅葡萄酒
首先介紹一下資料型別：
1.這個資料集包含了1599種紅酒，4898種白酒；
2.輸入資料特徵：
1 - fixed acidity
2 - volatile acidity
3 - citric acid
4 - residual sugar
5 - chlorides
6 - free sulfur dioxide
7 - total sulfur dioxide
8 - density
9 - pH
10 - sulphates
11 - alcohol
3.輸出變數：
12 - quality (score between 0 and 10)
import pandas as pd
#匯入資料
white = pd.read_csv("http://archive.ics.uci.edu/ml/machine-learning-databases/wine-quality/winequality-white.csv", sep=';')
red = pd.read_csv("http://archive.ics.uci.edu/ml/machine-learning-databases/wine-quality/winequality-red.csv", sep=';')#檢視白酒資訊
print white.info()<class 'pandas.core.frame.DataFrame'> RangeIndex: 4898 entries, 0 to 4897 Data columns (total 12 columns): fixed acidity 4898 non-null float64 volatile acidity 4898 non-null float64 citric acid 4898 non-null float64 residual sugar 4898 non-null float64 chlorides 4898 non-null float64 free sulfur dioxide 4898 non-null float64 total sulfur dioxide 4898 non-null float64 density 4898 non-null float64 pH 4898 non-null float64 sulphates 4898 non-null float64 alcohol 4898 non-null float64 quality 4898 non-null int64 dtypes: float64(11), int64(1) memory usage: 459.3 KB None
#檢視紅酒資訊
print red.info()<class 'pandas.core.frame.DataFrame'> RangeIndex: 1599 entries, 0 to 1598 Data columns (total 12 columns): fixed acidity 1599 non-null float64 volatile acidity 1599 non-null float64 citric acid 1599 non-null float64 residual sugar 1599 non-null float64 chlorides 1599 non-null float64 free sulfur dioxide 1599 non-null float64 total sulfur dioxide 1599 non-null float64 density 1599 non-null float64 pH 1599 non-null float64 sulphates 1599 non-null float64 alcohol 1599 non-null float64 quality 1599 non-null int64 dtypes: float64(11), int64(1) memory usage: 150.0 KB None
#檢視具體值
print red.head() fixed acidity volatile acidity citric acid residual sugar chlorides \
0 7.4 0.70 0.00 1.9 0.076
1 7.8 0.88 0.00 2.6 0.098
2 7.8 0.76 0.04 2.3 0.092
3 11.2 0.28 0.56 1.9 0.075
4 7.4 0.70 0.00 1.9 0.076
free sulfur dioxide total sulfur dioxide density pH sulphates \
0 11.0 34.0 0.9978 3.51 0.56
1 25.0 67.0 0.9968 3.20 0.68
2 15.0 54.0 0.9970 3.26 0.65
3 17.0 60.0 0.9980 3.16 0.58
4 11.0 34.0 0.9978 3.51 0.56
alcohol quality
0 9.4 5
1 9.8 5
2 9.8 5
3 9.8 6
4 9.4 5
#檢視各行統計資訊
print red.describe() fixed acidity volatile acidity citric acid residual sugar \
count 1599.000000 1599.000000 1599.000000 1599.000000
mean 8.319637 0.527821 0.270976 2.538806
std 1.741096 0.179060 0.194801 1.409928
min 4.600000 0.120000 0.000000 0.900000
25% 7.100000 0.390000 0.090000 1.900000
50% 7.900000 0.520000 0.260000 2.200000
75% 9.200000 0.640000 0.420000 2.600000
max 15.900000 1.580000 1.000000 15.500000
chlorides free sulfur dioxide total sulfur dioxide density \
count 1599.000000 1599.000000 1599.000000 1599.000000
mean 0.087467 15.874922 46.467792 0.996747
std 0.047065 10.460157 32.895324 0.001887
min 0.012000 1.000000 6.000000 0.990070
25% 0.070000 7.000000 22.000000 0.995600
50% 0.079000 14.000000 38.000000 0.996750
75% 0.090000 21.000000 62.000000 0.997835
max 0.611000 72.000000 289.000000 1.003690
pH sulphates alcohol quality
count 1599.000000 1599.000000 1599.000000 1599.000000
mean 3.311113 0.658149 10.422983 5.636023
std 0.154386 0.169507 1.065668 0.807569
min 2.740000 0.330000 8.400000 3.000000
25% 3.210000 0.550000 9.500000 5.000000
50% 3.310000 0.620000 10.200000 6.000000
75% 3.400000 0.730000 11.100000 6.000000
max 4.010000 2.000000 14.900000 8.000000
import numpy as np
#檢視是否有資料缺失
print np.any(red.isnull()==True)False
print np.any(white.isnull()==True)False
#視覺化資料
import matplotlib.pyplot as plt
fig,ax = plt.subplots(1,2)
ax[0].hist(red.alcohol, 10, facecolor='red', alpha=0.5, label="Red wine")
ax[1].hist(white.alcohol, 10, facecolor='white', ec="black", lw=0.5, alpha=0.5, label="White wine")
fig.subplots_adjust(left=0, right=1, bottom=0, top=0.5, hspace=0.05, wspace=1)
ax[0].set_ylim([0, 1000])
ax[0].set_xlabel("Alcohol in % Vol")
ax[0].set_ylabel("Frequency")
ax[1].set_xlabel("Alcohol in % Vol")
ax[1].set_ylabel("Frequency")
ax[0].legend(loc='best')
ax[1].legend(loc='best')
fig.suptitle("Distribution of Alcohol in % Vol")
plt.show()
我們可以從圖中看出紅酒和白酒的酒精濃度基本上9%左右。
#處理資料
#給我們的資料新增標籤
red['label'] = 1
white['label'] = 0
wines = red.append(white,ignore_index=True) #合併index順序import seaborn as sns
%matplotlib inline
corr = wines.corr() #計算協方差
sns.heatmap(corr,
xticklabels = corr.columns.values,
yticklabels = corr.columns.values)
sns.plt.show() #plt.show()---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
<ipython-input-18-5a30364cdd9b> in <module>()
5 xticklabels = corr.columns.values,
6 yticklabels = corr.columns.values)
----> 7 sns.plt.show()
AttributeError: 'module' object has no attribute 'plt'
這邊改成plt.show()就不會報錯了！
從圖中我們可以看到各個特徵之間的相關性，從中我們可以發現density跟residual sugar是正相關的，而跟alcohol是負相關的。
#劃分訓練集合測試集
from sklearn.model_selection import train_test_split
X = wines.iloc[:,0:11]
y = np.ravel(wines.label) #降成一維，類似np.flatten(),但是np.flatten是拷貝，而ravel是引用
#隨機劃分訓練集和測試集
#test_size:測試集佔比
#random_state:隨機種子，在需要重複試驗的時候，保證得到一組一樣的隨機數。
X_train,X_test,y_train,y_test = train_test_split(X,y,test_size=0.33, random_state=32)#標準化資料
from sklearn.preprocessing import StandardScaler
scaler = StandardScaler().fit(X_train)
X_train = scaler.transform(X_train)
X_test = scaler.transform(X_test)#使用keras模型化資料
from keras.models import Sequential
from keras.layers import Dense
model = Sequential()
#新增輸入層
model.add(Dense(12,activation='relu',
input_shape=(11,)))
#新增隱藏層
model.add(Dense(8,activation='relu'))
#新增輸出層
model.add(Dense(1,activation='sigmoid'))Using TensorFlow backend.
#檢視模型
#檢視輸出維度
print model.output_shape(None, 1)
#檢視整個模型
print model.summary()_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
dense_1 (Dense) (None, 12) 144
_________________________________________________________________
dense_2 (Dense) (None, 8) 104
_________________________________________________________________
dense_3 (Dense) (None, 1) 9
=================================================================
Total params: 257
Trainable params: 257
Non-trainable params: 0
_________________________________________________________________
None
#檢視模型引數
print model.get_weights()#模型的訓練
model.compile(loss='binary_crossentropy',
optimizer='adam',
metrics=['accuracy'])
#verbose = 1 檢視輸出過程
model.fit(X_train,y_train,epochs=30,batch_size=1,verbose=1)Epoch 1/30
4352/4352 [==============================] - 15s - loss: 0.1108 - acc: 0.9614
Epoch 2/30
4352/4352 [==============================] - 15s - loss: 0.0255 - acc: 0.9952
Epoch 3/30
4352/4352 [==============================] - 15s - loss: 0.0195 - acc: 0.9954
Epoch 4/30
4352/4352 [==============================] - 15s - loss: 0.0180 - acc: 0.9966
Epoch 5/30
4352/4352 [==============================] - 15s - loss: 0.0166 - acc: 0.9966
Epoch 6/30
4352/4352 [==============================] - 15s - loss: 0.0147 - acc: 0.9970
Epoch 7/30
4352/4352 [==============================] - 15s - loss: 0.0132 - acc: 0.9968
Epoch 8/30
4352/4352 [==============================] - 15s - loss: 0.0137 - acc: 0.9970
Epoch 9/30
4352/4352 [==============================] - 16s - loss: 0.0136 - acc: 0.9975
Epoch 10/30
4352/4352 [==============================] - 15s - loss: 0.0125 - acc: 0.9975
Epoch 11/30
4352/4352 [==============================] - 15s - loss: 0.0113 - acc: 0.9972
Epoch 12/30
4352/4352 [==============================] - 15s - loss: 0.0116 - acc: 0.9972
Epoch 13/30
4352/4352 [==============================] - 15s - loss: 0.0115 - acc: 0.9975
Epoch 14/30
4352/4352 [==============================] - 15s - loss: 0.0108 - acc: 0.9972
Epoch 15/30
4352/4352 [==============================] - 16s - loss: 0.0097 - acc: 0.9975
Epoch 16/30
4352/4352 [==============================] - 16s - loss: 0.0098 - acc: 0.9977
Epoch 17/30
4352/4352 [==============================] - 15s - loss: 0.0101 - acc: 0.9975
Epoch 18/30
4352/4352 [==============================] - 15s - loss: 0.0095 - acc: 0.9970
Epoch 19/30
4352/4352 [==============================] - 15s - loss: 0.0088 - acc: 0.9977
Epoch 20/30
4352/4352 [==============================] - 16s - loss: 0.0089 - acc: 0.9972
Epoch 21/30
4352/4352 [==============================] - 16s - loss: 0.0086 - acc: 0.9977
Epoch 22/30
4352/4352 [==============================] - 16s - loss: 0.0078 - acc: 0.9982
Epoch 23/30
4352/4352 [==============================] - 16s - loss: 0.0085 - acc: 0.9979
Epoch 24/30
4352/4352 [==============================] - 15s - loss: 0.0072 - acc: 0.9984
Epoch 25/30
4352/4352 [==============================] - 16s - loss: 0.0074 - acc: 0.9982
Epoch 26/30
4352/4352 [==============================] - 15s - loss: 0.0071 - acc: 0.9986
Epoch 27/30
4352/4352 [==============================] - 16s - loss: 0.0080 - acc: 0.9977
Epoch 28/30
4352/4352 [==============================] - 16s - loss: 0.0066 - acc: 0.9982
Epoch 29/30
4352/4352 [==============================] - 16s - loss: 0.0084 - acc: 0.9982
Epoch 30/30
4352/4352 [==============================] - 15s - loss: 0.0067 - acc: 0.9989
<keras.callbacks.History at 0x120530a90>
#預測結果
y_pred = model.predict(X_test)
print y_pred[:10][[ 2.14960589e-03]
[ 6.35436322e-07]
[ 1.82669051e-03]
[ 2.15678483e-07]
[ 1.00000000e+00]
[ 1.84882566e-07]
[ 1.13470778e-04]
[ 5.90343404e-07]
[ 2.01183035e-08]
[ 1.00000000e+00]]
print y_test[:10][0 0 0 0 1 0 0 0 0 1]
可以從上述結果可以看出，測試集的前十項結果跟我們預測的結果是一樣的。
#模型評估
score = model.evaluate(X_test,y_test,verbose=1)
#socre的兩個值分別代表損失(loss)和精準度(accuracy)
print score 960/2145 [============>.................] - ETA: 0s[0.030568929157742158, 0.99580419580419577]
#統計Precision、Recall、F1值
from sklearn.metrics import confusion_matrix,precision_score,recall_score,f1_score
y_pred = y_pred.astype(int) #轉化成整型
print confusion_matrix(y_test,y_pred)[[1633 0]
[ 230 282]]
#precision
precision = precision_score(y_test,y_pred)
print precision1.0
#Recall
recall = recall_score(y_test,y_pred)
print recall0.55078125
#F1 score
f1 = f1_score(y_test,y_pred)
print f10.710327455919
從上面的結果來看我們的Precision很高，但是我們的Recall值比較低。
對此我後面還會寫一篇blog來優化模型。