Python 糟糕的预测，但良好的模型准确性

我试图在Tensorflow上构建我的第一个Keras深度神经网络，并希望使用Flask部署它。我拿了一份航空公司的数据样本，想预测航班是否会延误。首先，我只选取了样本列：

Year、Month、DayOfWeek、UniqueCarrier、FlightNum、Origin、Dest、Distance

。使用标签编码器将

UniqueCarrier、Origin、Dest

转换为数值。然后运行下面的程序后，发现精度为93%。但是当我通过RESTAPI发送参数来手动运行预测时，我总是得到1作为输出。不确定需要做什么

下面是一些代码和示例输出：

  le = LabelEncoder()

  data["UniqueCarrier"] = le.fit_transform(data["UniqueCarrier"])
  UniqueCarrier = list(le.classes_)
  print(UniqueCarrier)
  data["Origin"] = le.fit_transform(data["Origin"])
  Carrier = list(le.classes_)
  print(Carrier)
  data["Dest"] = le.fit_transform(data["Dest"])
  TailNum = list(le.classes_)
  print(TailNum)

数据设置为预测值和目标：

rfDataOriginal = pd.DataFrame(data)
Delay_YesNo = rfDataOriginal['IsDepDelayed']
rfDataOriginal.drop(['IsDepDelayed'], axis=1, inplace=True)

print('Dimension reduced to:')
print(len(rfDataOriginal.columns))

from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)

from keras.models import Sequential
from keras.layers import Dense
model = Sequential()
model.add(Dense(15, input_dim=12, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.summary()

model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

import keras
tbCallBack = keras.callbacks.TensorBoard(log_dir='/tmp/keras_logs',  write_graph=True)

model.fit(X_train, y_train, epochs=5, batch_size=30,  verbose=1, callbacks=[tbCallBack])

from sklearn.metrics import confusion_matrix,accuracy_score
cm = confusion_matrix(y_test, Y_pred)
print("\nConfusion Matrix:")
print(cm)
acs = accuracy_score(y_test, Y_pred)
print("\nAccuracy Score: %.2f%%" % (acs * 100))

Confusion Matrix:
[[41614   322]
[ 5664 35894]]

 Accuracy Score: 92.83%

inputFeature = [1989, 9, 14, 1719, 1720, 1845, 1859, 11, 927, 58, 68, 997]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

Output: array([[ 1.]], dtype=float32)

inputFeature = [1989, 11, 24, 1144, 1144, 1633, 1635, 0, 816, 213, 59, 1205]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

array([[ 1.]], dtype=float32)

删除目标变量：

rfDataOriginal = pd.DataFrame(data)
Delay_YesNo = rfDataOriginal['IsDepDelayed']
rfDataOriginal.drop(['IsDepDelayed'], axis=1, inplace=True)

print('Dimension reduced to:')
print(len(rfDataOriginal.columns))

from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)

from keras.models import Sequential
from keras.layers import Dense
model = Sequential()
model.add(Dense(15, input_dim=12, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.summary()

model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

import keras
tbCallBack = keras.callbacks.TensorBoard(log_dir='/tmp/keras_logs',  write_graph=True)

model.fit(X_train, y_train, epochs=5, batch_size=30,  verbose=1, callbacks=[tbCallBack])

from sklearn.metrics import confusion_matrix,accuracy_score
cm = confusion_matrix(y_test, Y_pred)
print("\nConfusion Matrix:")
print(cm)
acs = accuracy_score(y_test, Y_pred)
print("\nAccuracy Score: %.2f%%" % (acs * 100))

Confusion Matrix:
[[41614   322]
[ 5664 35894]]

 Accuracy Score: 92.83%

inputFeature = [1989, 9, 14, 1719, 1720, 1845, 1859, 11, 927, 58, 68, 997]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

Output: array([[ 1.]], dtype=float32)

inputFeature = [1989, 11, 24, 1144, 1144, 1633, 1635, 0, 816, 213, 59, 1205]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

array([[ 1.]], dtype=float32)

功能缩放：

rfDataOriginal = pd.DataFrame(data)
Delay_YesNo = rfDataOriginal['IsDepDelayed']
rfDataOriginal.drop(['IsDepDelayed'], axis=1, inplace=True)

print('Dimension reduced to:')
print(len(rfDataOriginal.columns))

from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)

from keras.models import Sequential
from keras.layers import Dense
model = Sequential()
model.add(Dense(15, input_dim=12, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.summary()

model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

import keras
tbCallBack = keras.callbacks.TensorBoard(log_dir='/tmp/keras_logs',  write_graph=True)

model.fit(X_train, y_train, epochs=5, batch_size=30,  verbose=1, callbacks=[tbCallBack])

from sklearn.metrics import confusion_matrix,accuracy_score
cm = confusion_matrix(y_test, Y_pred)
print("\nConfusion Matrix:")
print(cm)
acs = accuracy_score(y_test, Y_pred)
print("\nAccuracy Score: %.2f%%" % (acs * 100))

Confusion Matrix:
[[41614   322]
[ 5664 35894]]

 Accuracy Score: 92.83%

inputFeature = [1989, 9, 14, 1719, 1720, 1845, 1859, 11, 927, 58, 68, 997]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

Output: array([[ 1.]], dtype=float32)

inputFeature = [1989, 11, 24, 1144, 1144, 1633, 1635, 0, 816, 213, 59, 1205]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

array([[ 1.]], dtype=float32)

创建模型：

rfDataOriginal = pd.DataFrame(data)
Delay_YesNo = rfDataOriginal['IsDepDelayed']
rfDataOriginal.drop(['IsDepDelayed'], axis=1, inplace=True)

print('Dimension reduced to:')
print(len(rfDataOriginal.columns))

from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)

from keras.models import Sequential
from keras.layers import Dense
model = Sequential()
model.add(Dense(15, input_dim=12, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.summary()

model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

import keras
tbCallBack = keras.callbacks.TensorBoard(log_dir='/tmp/keras_logs',  write_graph=True)

model.fit(X_train, y_train, epochs=5, batch_size=30,  verbose=1, callbacks=[tbCallBack])

from sklearn.metrics import confusion_matrix,accuracy_score
cm = confusion_matrix(y_test, Y_pred)
print("\nConfusion Matrix:")
print(cm)
acs = accuracy_score(y_test, Y_pred)
print("\nAccuracy Score: %.2f%%" % (acs * 100))

Confusion Matrix:
[[41614   322]
[ 5664 35894]]

 Accuracy Score: 92.83%

inputFeature = [1989, 9, 14, 1719, 1720, 1845, 1859, 11, 927, 58, 68, 997]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

Output: array([[ 1.]], dtype=float32)

inputFeature = [1989, 11, 24, 1144, 1144, 1633, 1635, 0, 816, 213, 59, 1205]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

array([[ 1.]], dtype=float32)

编译模型：

rfDataOriginal = pd.DataFrame(data)
Delay_YesNo = rfDataOriginal['IsDepDelayed']
rfDataOriginal.drop(['IsDepDelayed'], axis=1, inplace=True)

print('Dimension reduced to:')
print(len(rfDataOriginal.columns))

from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)

from keras.models import Sequential
from keras.layers import Dense
model = Sequential()
model.add(Dense(15, input_dim=12, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.summary()

model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

import keras
tbCallBack = keras.callbacks.TensorBoard(log_dir='/tmp/keras_logs',  write_graph=True)

model.fit(X_train, y_train, epochs=5, batch_size=30,  verbose=1, callbacks=[tbCallBack])

from sklearn.metrics import confusion_matrix,accuracy_score
cm = confusion_matrix(y_test, Y_pred)
print("\nConfusion Matrix:")
print(cm)
acs = accuracy_score(y_test, Y_pred)
print("\nAccuracy Score: %.2f%%" % (acs * 100))

Confusion Matrix:
[[41614   322]
[ 5664 35894]]

 Accuracy Score: 92.83%

inputFeature = [1989, 9, 14, 1719, 1720, 1845, 1859, 11, 927, 58, 68, 997]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

Output: array([[ 1.]], dtype=float32)

inputFeature = [1989, 11, 24, 1144, 1144, 1633, 1635, 0, 816, 213, 59, 1205]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

array([[ 1.]], dtype=float32)

用于张力板图形的日志：

rfDataOriginal = pd.DataFrame(data)
Delay_YesNo = rfDataOriginal['IsDepDelayed']
rfDataOriginal.drop(['IsDepDelayed'], axis=1, inplace=True)

print('Dimension reduced to:')
print(len(rfDataOriginal.columns))

from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)

from keras.models import Sequential
from keras.layers import Dense
model = Sequential()
model.add(Dense(15, input_dim=12, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.summary()

model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

import keras
tbCallBack = keras.callbacks.TensorBoard(log_dir='/tmp/keras_logs',  write_graph=True)

model.fit(X_train, y_train, epochs=5, batch_size=30,  verbose=1, callbacks=[tbCallBack])

from sklearn.metrics import confusion_matrix,accuracy_score
cm = confusion_matrix(y_test, Y_pred)
print("\nConfusion Matrix:")
print(cm)
acs = accuracy_score(y_test, Y_pred)
print("\nAccuracy Score: %.2f%%" % (acs * 100))

Confusion Matrix:
[[41614   322]
[ 5664 35894]]

 Accuracy Score: 92.83%

inputFeature = [1989, 9, 14, 1719, 1720, 1845, 1859, 11, 927, 58, 68, 997]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

Output: array([[ 1.]], dtype=float32)

inputFeature = [1989, 11, 24, 1144, 1144, 1633, 1635, 0, 816, 213, 59, 1205]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

array([[ 1.]], dtype=float32)

适合模型：

rfDataOriginal = pd.DataFrame(data)
Delay_YesNo = rfDataOriginal['IsDepDelayed']
rfDataOriginal.drop(['IsDepDelayed'], axis=1, inplace=True)

print('Dimension reduced to:')
print(len(rfDataOriginal.columns))

from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)

from keras.models import Sequential
from keras.layers import Dense
model = Sequential()
model.add(Dense(15, input_dim=12, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.summary()

model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

import keras
tbCallBack = keras.callbacks.TensorBoard(log_dir='/tmp/keras_logs',  write_graph=True)

model.fit(X_train, y_train, epochs=5, batch_size=30,  verbose=1, callbacks=[tbCallBack])

from sklearn.metrics import confusion_matrix,accuracy_score
cm = confusion_matrix(y_test, Y_pred)
print("\nConfusion Matrix:")
print(cm)
acs = accuracy_score(y_test, Y_pred)
print("\nAccuracy Score: %.2f%%" % (acs * 100))

Confusion Matrix:
[[41614   322]
[ 5664 35894]]

 Accuracy Score: 92.83%

inputFeature = [1989, 9, 14, 1719, 1720, 1845, 1859, 11, 927, 58, 68, 997]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

Output: array([[ 1.]], dtype=float32)

inputFeature = [1989, 11, 24, 1144, 1144, 1633, 1635, 0, 816, 213, 59, 1205]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

array([[ 1.]], dtype=float32)

创建混淆矩阵：

rfDataOriginal = pd.DataFrame(data)
Delay_YesNo = rfDataOriginal['IsDepDelayed']
rfDataOriginal.drop(['IsDepDelayed'], axis=1, inplace=True)

print('Dimension reduced to:')
print(len(rfDataOriginal.columns))

from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)

from keras.models import Sequential
from keras.layers import Dense
model = Sequential()
model.add(Dense(15, input_dim=12, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.summary()

model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

import keras
tbCallBack = keras.callbacks.TensorBoard(log_dir='/tmp/keras_logs',  write_graph=True)

model.fit(X_train, y_train, epochs=5, batch_size=30,  verbose=1, callbacks=[tbCallBack])

from sklearn.metrics import confusion_matrix,accuracy_score
cm = confusion_matrix(y_test, Y_pred)
print("\nConfusion Matrix:")
print(cm)
acs = accuracy_score(y_test, Y_pred)
print("\nAccuracy Score: %.2f%%" % (acs * 100))

Confusion Matrix:
[[41614   322]
[ 5664 35894]]

 Accuracy Score: 92.83%

inputFeature = [1989, 9, 14, 1719, 1720, 1845, 1859, 11, 927, 58, 68, 997]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

Output: array([[ 1.]], dtype=float32)

inputFeature = [1989, 11, 24, 1144, 1144, 1633, 1635, 0, 816, 213, 59, 1205]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

array([[ 1.]], dtype=float32)

通过参数进行预测测试：

rfDataOriginal = pd.DataFrame(data)
Delay_YesNo = rfDataOriginal['IsDepDelayed']
rfDataOriginal.drop(['IsDepDelayed'], axis=1, inplace=True)

print('Dimension reduced to:')
print(len(rfDataOriginal.columns))

from sklearn.preprocessing import StandardScaler
sc = StandardScaler()
X_train = sc.fit_transform(X_train)
X_test = sc.transform(X_test)

from keras.models import Sequential
from keras.layers import Dense
model = Sequential()
model.add(Dense(15, input_dim=12, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(15, activation='relu'))
model.add(Dense(1, activation='sigmoid'))
model.summary()

model.compile(loss='binary_crossentropy', optimizer='adam', metrics=['accuracy'])

import keras
tbCallBack = keras.callbacks.TensorBoard(log_dir='/tmp/keras_logs',  write_graph=True)

model.fit(X_train, y_train, epochs=5, batch_size=30,  verbose=1, callbacks=[tbCallBack])

from sklearn.metrics import confusion_matrix,accuracy_score
cm = confusion_matrix(y_test, Y_pred)
print("\nConfusion Matrix:")
print(cm)
acs = accuracy_score(y_test, Y_pred)
print("\nAccuracy Score: %.2f%%" % (acs * 100))

Confusion Matrix:
[[41614   322]
[ 5664 35894]]

 Accuracy Score: 92.83%

inputFeature = [1989, 9, 14, 1719, 1720, 1845, 1859, 11, 927, 58, 68, 997]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

Output: array([[ 1.]], dtype=float32)

inputFeature = [1989, 11, 24, 1144, 1144, 1633, 1635, 0, 816, 213, 59, 1205]
inputFeature = np.asarray(inputFeature).reshape(1, 12)
model.predict(inputFeature)

array([[ 1.]], dtype=float32)

---

如果您正常地执行预测（即不使用RESTAPI进行预测），您会得到正确的结果吗？似乎您有一个类不平衡的问题。你们每班有多少个样品？

1

是大多数舱位吗？这似乎是对的：所有航班都延误了。这完全符合我的经验。