python - 如何应用经过训练的神经网络将输出写入 csv 文件？

Question

我有以下数据集data_num，我已在该数据集上训练了神经网络，这是data_num的示例:

    A   B   C   D  Label1
0  95  91   3  10       9
1  91  95   4   3       9
2  68  65  31  31       6
3  50  43  51  58       4
4   8   4  93  97       0
5  16  20  81  90       1
6  75  79  28  26       7
7  74  76  27  22       7
8  45  46  56  57       4
9   5   7  97  93       0

这是完整的代码:

import keras
from keras.models import Sequential
from keras.layers import Dense, Dropout, Activation
from keras.optimizers import SGD
import pandas as pd
import numpy as np
import matplotlib
from matplotlib import style
import matplotlib.pyplot as plt
from sklearn.model_selection import train_test_split
from IPython.core.display import display
from sklearn.preprocessing import MinMaxScaler

matplotlib.style.use('ggplot')

data_num = pd.read_csv('mult_test.csv')

scaler = MinMaxScaler(feature_range=(0, 1))
features = data_num.drop(['Label1'], axis=1, errors='ignore')
features = pd.DataFrame(scaler.fit_transform(features))
scale_num_data = pd.concat([data_num['Label1'], features], axis=1)


dtrain, dtest = train_test_split(scale_num_data, test_size=0.25, random_state=570)
X = dtrain.drop(['Label1'], axis=1, errors='ignore')
y = dtrain['Label1']
Xtest = dtest.drop(['Label1'], axis=1, errors='ignore')
ytest = dtest['Label1']


model = Sequential([
    Dense(10, input_shape=(4, ), activation='relu'),
    Dense(32, activation='relu'),
    Dense(10, activation='softmax')
])

model.summary()
model.compile(optimizer='rmsprop', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
model.fit(X, y, epochs=10, batch_size=10, shuffle=True)


scores = model.evaluate(Xtest, ytest, batch_size=5)
print("%s: %.2f%%" % (model.metrics_names[1], scores[1]*100))

现在网络已经训练完毕，我想将代码应用到新数据集 predict_num 并将 NN 输出打印到新列“预测”中。以下是新数据集的示例:

  Stock   A   B    C   D  Prediction
0  AMCX  46  43   57  52           
1  ABAC  83  86   11  18           
2  AKAM  55  52   45  43           
3   ACW  96  99    9   8           
4  AOLS  46  43   54  52           
5  ABAX   9   9  100  95           
6  AMTX   9   1   91  97           
7  ABMC  73  79   29  25           
8   ALE  58  56   50  44           
9  AMAT   8   1   98  92           

predict_num 还有一个额外的列“Stock”，因此我只想将特征指定为列 [A、B、C、D]，并用输出填充“预测”列神经网络的。

最终的数据集应如下所示:

  Stock   A   B    C   D  Prediction
0  AMCX  46  43   57  52           4
1  ABAC  83  86   11  18           8
2  AKAM  55  52   45  43           5
3   ACW  96  99    9   8           9
4  AOLS  46  43   54  52           4
5  ABAX   9   9  100  95           0
6  AMTX   9   1   91  97           0
7  ABMC  73  79   29  25           7
8   ALE  58  56   50  44           5
9  AMAT   8   1   98  92           0

非常感谢您的帮助。

Answer 1

使用model.predict

.... your code ....
model.summary()
model.compile(optimizer='rmsprop', loss='sparse_categorical_crossentropy', metrics=['accuracy'])
model.fit(X, y, epochs=10, batch_size=10, shuffle=True)

pred = model.predict(xtest)
xtest["prediciton"] = pred
xtest.to_csv("my_new_file.csv")

:)