Question

我正在做一些练习神经网络，并且只停留在一个问题上。我的网络无法预测出正确的结果，即使它说训练得分为97％。

这是我的代码：

# Import `datasets` from `sklearn`
from sklearn import datasets
import pandas as pd
from sklearn.model_selection import train_test_split
# Import `train_test_split`
from sklearn.model_selection  import train_test_split
from sklearn.preprocessing import StandardScaler
from sklearn.neural_network import MLPClassifier


# Load in the `digits` data
from sklearn.preprocessing import scale

iris = datasets.load_iris()

# split the data up - 3/4 for training, 1/4 for testing
data_train, data_test, name_train, name_test = train_test_split(iris.data, 
iris.target, test_size=0.25, random_state=0)

# Number of training features
# n_samples, n_features = data_train.shape

scaler = StandardScaler()
scaler.fit(data_train)
params_train_scaled = scaler.transform(data_train)
params_test_scaled = scaler.transform(data_test)

# 1 hidden layer, same size as the input layer
mlp = MLPClassifier(
    solver='lbfgs',
    hidden_layer_sizes=(iris.data.shape[1], ),
    random_state=0)
mlp.fit(params_train_scaled, name_train)

print(name_train)
print('Train score: %.3g' % mlp.score(params_train_scaled, name_train))
print('Test Score: %.3g' % mlp.score(params_test_scaled, name_test))
print

test_val = [[5.1, 3.5, 1.4, 0.2]]

print(mlp.predict(test_val))

我想这就是我如何扩展训练和测试数据的方法，但是我不确定...

我得到的输出是：

火车成绩：1

测试成绩：0.974

但是，预测值应为0，而不是1。

感谢您的帮助。

Answer 1

由于您缩放了训练数据，因此还应该缩放测试数据：

user_id|sub_category_id|category_id|is_removed
5      |201            |2          |false

Sklearn神经网络问题

1 个答案: