• sklearn中神经网络API
• sklearn.neural_network.MLPClassifier
• sklearn.neural_network.MLPRegressor

使用pandas导入数据

from sklearn.neural_network import MLPRegressor
import pandas as pd
import numpy as np
data_tr = pd.read_csv(r'./data/BPdata_tr.txt')   # 导入数据
data_te = pd.read_csv(r'./data/BPdata_te.txt')

查看data_tr

data_tr
           x1        x2         y
0    0.065735  0.044223  0.006277
1    0.049985  0.334657  0.114494
2    0.541525  0.027610  0.294011
3    0.038182  0.284811  0.082575
4    0.262401  0.198246  0.108156
..        ...       ...       ...
495  0.343548  0.500284  0.368309
496  0.259262  0.558471  0.379106
497  0.297697  0.123911  0.103977
498  0.507976  0.304445  0.350726
499  0.396284  0.607128  0.525645
[500 rows x 3 columns]

调用API训练模型

model = MLPRegressor(hidden_layer_sizes=(10,), random_state=10,learning_rate_init=0.1)  # BP神经网络回归模型
model.fit(data_tr.iloc[:,:2],data_tr.iloc[:,2])  # 训练模型
pre = model.predict(data_te.iloc[:,:2])  # 模型预测
np.abs(data_te.iloc[:,2]-pre).mean()  # 模型评价

查看模型预测值与实际差值绝对值的平均值

np.abs(data_te.iloc[:,2]-pre).mean()  # 模型评价
0.037916438908481044

• 总结：API调用较容易，但重点在于对模型中算法的理解，想要调试好模型，需调试好参数。