Question

输入：语音的50种深层特征

大小：20000+个样本

求解器：亚当

输出类别：10（数字0至9）

隐藏层和神经元：3和100,100,100

代码：

X_train, X_test, y_train, y_test = train_test_split(Combineddatafeatures,
Combineddatalabels, test_size=0.1, random_state=1)

clf = MLPClassifier(solver='adam',hidden_layer_sizes=(100,100,100))

clf.fit(np.array(X_train), np.array(y_train)) 

predicted_values = clf.predict(X_test)

from sklearn.metrics import accuracy_score
score = accuracy_score(y_test,predicted_values)

print(score)

当前精度： 0.5693142575234337

如何提高样品的准确性（20262，50形状）（？）通过微调

隐藏层数
隐藏层中的神经元

Answer 1

进行参数调整的一种方法是使用sklearn的GridSearchCV方法：link

它将测试您将给出的所有参数组合并输出最佳组合。

from sklearn.model_selection import GridSearchCV

X_train, X_test, y_train, y_test = train_test_split(Combineddatafeatures,
Combineddatalabels, test_size=0.1, random_state=1)

clf = MLPClassifier(solver='adam',hidden_layer_sizes=(100,100,100))

params = {
    'hidden_layer_sizes' : [(100, 100, 100), (125, 125, 125)] # Every combination you want to try
}

gscv = GridSearchCV(clf, params, verbose=1)

gscv.fit(np.array(X_train), np.array(y_train))
print(gscv.best_params_) 

predicted_values = gscv.predict(X_test)

from sklearn.metrics import accuracy_score
score = accuracy_score(y_test,predicted_values)

print(score)

MLPClassifier模型精度可通过20000个样本进行微调

1 个答案: