\u6570\u636e\u6807\u51c6\u5316<\/strong><\/h2>\nscaler = StandardScaler()<\/p>\n
X_train = scaler.fit_transform(X_train)<\/p>\n
X_test = scaler.transform(X_test)<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u4e09\u3001\u5b9a\u4e49\u6a21\u578b<\/h2>\n<\/p>\n
\u5728\u5b8c\u6210\u6570\u636e\u51c6\u5907\u540e\uff0c\u9700\u8981\u5b9a\u4e49\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u3002\u5728Keras\u4e2d\uff0c\u53ef\u4ee5\u901a\u8fc7Sequential API\u6216Functional API\u5b9a\u4e49\u6a21\u578b\u3002<\/p>\n<\/p>\n
1. Sequential API<\/h3>\n<\/p>\n
Sequential API\u9002\u5408\u6784\u5efa\u7ebf\u6027\u5806\u53e0\u7684\u6a21\u578b\uff0c\u5373\u6bcf\u4e00\u5c42\u7684\u8f93\u51fa\u4f5c\u4e3a\u4e0b\u4e00\u5c42\u7684\u8f93\u5165\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u793a\u4f8b\uff1a<\/p>\n<\/p>\n
from keras.models import Sequential<\/p>\nfrom keras.layers import Dense<\/p>\n
\u5b9a\u4e49\u4e00\u4e2a\u987a\u5e8f\u6a21\u578b<\/strong><\/h2>\nmodel = Sequential()<\/p>\n
\u6dfb\u52a0\u8f93\u5165\u5c42\u548c\u7b2c\u4e00\u4e2a\u9690\u85cf\u5c42<\/strong><\/h2>\nmodel.add(Dense(units=64, activation='relu', input_dim=20))<\/p>\n
\u6dfb\u52a0\u7b2c\u4e8c\u4e2a\u9690\u85cf\u5c42<\/strong><\/h2>\nmodel.add(Dense(units=64, activation='relu'))<\/p>\n
\u6dfb\u52a0\u8f93\u51fa\u5c42<\/strong><\/h2>\nmodel.add(Dense(units=1, activation='sigmoid'))<\/p>\n
\u7f16\u8bd1\u6a21\u578b<\/strong><\/h2>\nmodel.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])<\/p>\n
<\/code><\/pre>\n<\/p>\n
2. Functional API<\/h3>\n<\/p>\n
Functional API\u9002\u5408\u6784\u5efa\u590d\u6742\u7684\u6a21\u578b\uff0c\u5982\u591a\u8f93\u5165\u591a\u8f93\u51fa\u6a21\u578b\u3001\u5171\u4eab\u5c42\u7b49\u3002\u4ee5\u4e0b\u662f\u4e00\u4e2a\u7b80\u5355\u7684\u793a\u4f8b\uff1a<\/p>\n<\/p>\n
from keras.models import Model<\/p>\nfrom keras.layers import Input, Dense<\/p>\n
\u5b9a\u4e49\u8f93\u5165\u5c42<\/strong><\/h2>\ninputs = Input(shape=(20,))<\/p>\n
\u5b9a\u4e49\u7b2c\u4e00\u4e2a\u9690\u85cf\u5c42<\/strong><\/h2>\nx = Dense(units=64, activation='relu')(inputs)<\/p>\n
\u5b9a\u4e49\u7b2c\u4e8c\u4e2a\u9690\u85cf\u5c42<\/strong><\/h2>\nx = Dense(units=64, activation='relu')(x)<\/p>\n
\u5b9a\u4e49\u8f93\u51fa\u5c42<\/strong><\/h2>\noutputs = Dense(units=1, activation='sigmoid')(x)<\/p>\n
\u6784\u5efa\u6a21\u578b<\/strong><\/h2>\nmodel = Model(inputs=inputs, outputs=outputs)<\/p>\n
\u7f16\u8bd1\u6a21\u578b<\/strong><\/h2>\nmodel.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u56db\u3001\u8bad\u7ec3\u6a21\u578b<\/h2>\n<\/p>\n
\u5728\u5b9a\u4e49\u597d\u6a21\u578b\u540e\uff0c\u9700\u8981\u4f7f\u7528\u8bad\u7ec3\u6570\u636e\u5bf9\u6a21\u578b\u8fdb\u884c\u8bad\u7ec3\u3002\u8bad\u7ec3\u7684\u4e3b\u8981\u6b65\u9aa4\u5305\u62ec\uff1a<\/p>\n<\/p>\n
1. \u8bbe\u7f6e\u8d85\u53c2\u6570<\/h3>\n<\/p>\n
\u8d85\u53c2\u6570\u662f\u6307\u6a21\u578b\u8bad\u7ec3\u8fc7\u7a0b\u4e2d\u7684\u53c2\u6570\uff0c\u5982\u5b66\u4e60\u7387\u3001\u6279\u6b21\u5927\u5c0f\u3001\u8bad\u7ec3\u8f6e\u6570\u7b49\u3002\u5408\u7406\u8bbe\u7f6e\u8d85\u53c2\u6570\u53ef\u4ee5\u63d0\u9ad8\u6a21\u578b\u7684\u6027\u80fd\u3002<\/p>\n<\/p>\n
2. \u8bad\u7ec3\u6a21\u578b<\/h3>\n<\/p>\n
\u4f7f\u7528\u8bad\u7ec3\u6570\u636e\u5bf9\u6a21\u578b\u8fdb\u884c\u8bad\u7ec3\uff0c\u5e76\u5728\u9a8c\u8bc1\u6570\u636e\u4e0a\u8bc4\u4f30\u6a21\u578b\u7684\u6027\u80fd\u3002\u8bad\u7ec3\u8fc7\u7a0b\u53ef\u4ee5\u901a\u8fc7\u56de\u8c03\u51fd\u6570\u8fdb\u884c\u76d1\u63a7\uff0c\u5982EarlyStopping\u3001ModelCheckpoint\u7b49\u3002<\/p>\n<\/p>\n
\u793a\u4f8b\u4ee3\u7801\uff1a<\/h3>\n<\/p>\n
from keras.callbacks import EarlyStopping, ModelCheckpoint<\/p>\n\u8bbe\u7f6e\u56de\u8c03\u51fd\u6570<\/strong><\/h2>\nearly_stopping = EarlyStopping(monitor='val_loss', patience=10)<\/p>\n
model_checkpoint = ModelCheckpoint('best_model.h5', save_best_only=True)<\/p>\n
\u8bad\u7ec3\u6a21\u578b<\/strong><\/h2>\nhistory = model.fit(X_train, y_train, <\/p>\n
epochs=100, <\/p>\n
batch_size=32, <\/p>\n
validation_split=0.2, <\/p>\n
callbacks=[early_stopping, model_checkpoint])<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u4e94\u3001\u8bc4\u4f30\u6a21\u578b<\/h2>\n<\/p>\n
\u5728\u8bad\u7ec3\u5b8c\u6210\u540e\uff0c\u9700\u8981\u4f7f\u7528\u6d4b\u8bd5\u6570\u636e\u5bf9\u6a21\u578b\u8fdb\u884c\u8bc4\u4f30\uff0c\u4ee5\u9a8c\u8bc1\u6a21\u578b\u7684\u6cdb\u5316\u80fd\u529b\u3002\u8bc4\u4f30\u7684\u4e3b\u8981\u6b65\u9aa4\u5305\u62ec\uff1a<\/p>\n<\/p>\n
1. \u6a21\u578b\u9884\u6d4b<\/h3>\n<\/p>\n
\u4f7f\u7528\u6d4b\u8bd5\u6570\u636e\u8fdb\u884c\u9884\u6d4b\uff0c\u5e76\u8ba1\u7b97\u6a21\u578b\u7684\u6027\u80fd\u6307\u6807\uff0c\u5982\u51c6\u786e\u7387\u3001\u7cbe\u786e\u7387\u3001\u53ec\u56de\u7387\u7b49\u3002<\/p>\n<\/p>\n
2. \u6a21\u578b\u8bc4\u4f30<\/h3>\n<\/p>\n
\u901a\u8fc7\u6df7\u6dc6\u77e9\u9635\u3001ROC\u66f2\u7ebf\u7b49\u53ef\u89c6\u5316\u5de5\u5177\uff0c\u5bf9\u6a21\u578b\u7684\u6027\u80fd\u8fdb\u884c\u5168\u9762\u8bc4\u4f30\u3002<\/p>\n<\/p>\n
\u793a\u4f8b\u4ee3\u7801\uff1a<\/h3>\n<\/p>\n
from sklearn.metrics import accuracy_score, confusion_matrix, roc_curve, auc<\/p>\nimport matplotlib.pyplot as plt<\/p>\n
\u6a21\u578b\u9884\u6d4b<\/strong><\/h2>\ny_pred = model.predict(X_test)<\/p>\n
y_pred_classes = (y_pred > 0.5).astype(int)<\/p>\n
\u8ba1\u7b97\u51c6\u786e\u7387<\/strong><\/h2>\naccuracy = accuracy_score(y_test, y_pred_classes)<\/p>\n
print('Accuracy:', accuracy)<\/p>\n
\u7ed8\u5236\u6df7\u6dc6\u77e9\u9635<\/strong><\/h2>\nconf_matrix = confusion_matrix(y_test, y_pred_classes)<\/p>\n
plt.figure(figsize=(8, 6))<\/p>\n
sns.heatmap(conf_matrix, annot=True, fmt='d', cmap='Blues')<\/p>\n
plt.xlabel('Predicted')<\/p>\n
plt.ylabel('True')<\/p>\n
plt.show()<\/p>\n
\u7ed8\u5236 ROC \u66f2\u7ebf<\/strong><\/h2>\nfpr, tpr, _ = roc_curve(y_test, y_pred)<\/p>\n
roc_auc = auc(fpr, tpr)<\/p>\n
plt.figure(figsize=(8, 6))<\/p>\n
plt.plot(fpr, tpr, color='blue', label='ROC curve (area = %0.2f)' % roc_auc)<\/p>\n
plt.plot([0, 1], [0, 1], color='gray', linestyle='--')<\/p>\n
plt.xlabel('False Positive Rate')<\/p>\n
plt.ylabel('True Positive Rate')<\/p>\n
plt.title('Receiver Operating Characteristic (ROC) Curve')<\/p>\n
plt.legend(loc='lower right')<\/p>\n
plt.show()<\/p>\n
<\/code><\/pre>\n<\/p>\n
\u901a\u8fc7\u4ee5\u4e0a\u6b65\u9aa4\uff0c\u6211\u4eec\u53ef\u4ee5\u5728Python\u4e2d\u6784\u5efa\u3001\u8bad\u7ec3\u548c\u8bc4\u4f30\u4e00\u4e2a\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u3002\u65e0\u8bba\u662f\u521d\u5b66\u8005\u8fd8\u662f\u6709\u7ecf\u9a8c\u7684\u5f00\u53d1\u8005\uff0c\u90fd\u53ef\u4ee5\u6839\u636e\u5b9e\u9645\u9700\u6c42\u9009\u62e9\u5408\u9002\u7684\u5e93\u548c\u65b9\u6cd5\uff0c\u6784\u5efa\u51fa\u6027\u80fd\u4f18\u5f02\u7684\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u3002<\/p>\n<\/p>\n
\u76f8\u5173\u95ee\u7b54FAQs\uff1a<\/strong><\/h2>\n \u5728Python\u4e2d\u4f7f\u7528\u795e\u7ecf\u7f51\u7edc\u9700\u8981\u54ea\u4e9b\u57fa\u672c\u7684\u5e93\u548c\u5de5\u5177\uff1f<\/strong>
\u5728Python\u4e2d\u6784\u5efa\u795e\u7ecf\u7f51\u7edc\uff0c\u901a\u5e38\u9700\u8981\u4e00\u4e9b\u5173\u952e\u7684\u5e93\uff0c\u5982TensorFlow\u548cKeras\uff0cPyTorch\u7b49\u3002\u8fd9\u4e9b\u5e93\u63d0\u4f9b\u4e86\u4e30\u5bcc\u7684\u529f\u80fd\u548c\u7075\u6d3b\u6027\uff0c\u4f7f\u5f97\u8bbe\u8ba1\u548c\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\u53d8\u5f97\u66f4\u52a0\u7b80\u5355\u3002\u6b64\u5916\uff0cNumPy\u548cPandas\u7b49\u5e93\u4e5f\u5f88\u91cd\u8981\uff0c\u56e0\u4e3a\u5b83\u4eec\u7528\u4e8e\u6570\u636e\u5904\u7406\u548c\u5206\u6790\u3002\u786e\u4fdd\u5b89\u88c5\u8fd9\u4e9b\u5e93\u540e\uff0c\u53ef\u4ee5\u5f00\u59cb\u6784\u5efa\u548c\u8bad\u7ec3\u4f60\u7684\u795e\u7ecf\u7f51\u7edc\u6a21\u578b\u3002<\/p>\n\u6211\u5982\u4f55\u9009\u62e9\u9002\u5408\u6211\u7684\u9879\u76ee\u7684\u795e\u7ecf\u7f51\u7edc\u67b6\u6784\uff1f<\/strong>
\u9009\u62e9\u5408\u9002\u7684\u795e\u7ecf\u7f51\u7edc\u67b6\u6784\u901a\u5e38\u53d6\u51b3\u4e8e\u9879\u76ee\u7684\u5177\u4f53\u9700\u6c42\u3002\u5bf9\u4e8e\u56fe\u50cf\u5904\u7406\u4efb\u52a1\uff0c\u5377\u79ef\u795e\u7ecf\u7f51\u7edc\uff08CNN\uff09\u5e38\u5e38\u662f\u9996\u9009\uff1b\u800c\u5904\u7406\u5e8f\u5217\u6570\u636e\u65f6\uff0c\u5faa\u73af\u795e\u7ecf\u7f51\u7edc\uff08RNN\uff09\u6216\u957f\u77ed\u671f\u8bb0\u5fc6\u7f51\u7edc\uff08LSTM\uff09\u53ef\u80fd\u66f4\u4e3a\u5408\u9002\u3002\u4e86\u89e3\u4f60\u7684\u6570\u636e\u7c7b\u578b\u548c\u4efb\u52a1\u76ee\u6807\u662f\u9009\u62e9\u5408\u9002\u67b6\u6784\u7684\u5173\u952e\u3002\u6b64\u5916\uff0c\u53ef\u4ee5\u53c2\u8003\u4e00\u4e9b\u73b0\u6210\u7684\u6a21\u578b\u7ed3\u6784\uff0c\u5982ResNet\u6216VGG\uff0c\u4ee5\u6b64\u4f5c\u4e3a\u8d77\u70b9\u3002<\/p>\n\u5728\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\u65f6\uff0c\u6211\u5e94\u8be5\u6ce8\u610f\u54ea\u4e9b\u8d85\u53c2\u6570\u7684\u8c03\u6574\uff1f<\/strong>
\u8bad\u7ec3\u795e\u7ecf\u7f51\u7edc\u65f6\uff0c\u8d85\u53c2\u6570\u7684\u8c03\u6574\u81f3\u5173\u91cd\u8981\uff0c\u5305\u62ec\u5b66\u4e60\u7387\u3001\u6279\u91cf\u5927\u5c0f\u3001\u8bad\u7ec3\u8f6e\u6570\u548c\u7f51\u7edc\u5c42\u6570\u7b49\u3002\u5b66\u4e60\u7387\u5f71\u54cd\u6a21\u578b\u7684\u6536\u655b\u901f\u5ea6\uff0c\u6279\u91cf\u5927\u5c0f\u4f1a\u5f71\u54cd\u6a21\u578b\u7684\u7a33\u5b9a\u6027\u548c\u8bad\u7ec3\u65f6\u95f4\u3002\u9002\u5f53\u7684\u8bad\u7ec3\u8f6e\u6570\u53ef\u4ee5\u786e\u4fdd\u6a21\u578b\u5145\u5206\u5b66\u4e60\uff0c\u800c\u7f51\u7edc\u5c42\u6570\u548c\u6bcf\u5c42\u7684\u795e\u7ecf\u5143\u6570\u91cf\u5219\u51b3\u5b9a\u4e86\u6a21\u578b\u7684\u590d\u6742\u6027\u3002\u4f7f\u7528\u4ea4\u53c9\u9a8c\u8bc1\u548c\u7f51\u683c\u641c\u7d22\u7b49\u65b9\u6cd5\uff0c\u53ef\u4ee5\u5e2e\u52a9\u627e\u5230\u6700\u4f73\u7684\u8d85\u53c2\u6570\u7ec4\u5408\u3002<\/p>\n","protected":false},"excerpt":{"rendered":"\u5728Python\u4e2d\u4f7f\u7528\u795e\u7ecf\u7f51\u7edc\u7684\u6838\u5fc3\u8981\u70b9\u6709\uff1a\u9009\u62e9\u9002\u5f53\u7684\u5e93\u3001\u51c6\u5907\u6570\u636e\u3001\u5b9a\u4e49\u6a21\u578b\u3001\u8bad\u7ec3\u6a21\u578b\u3001\u8bc4\u4f30\u6a21\u578b\u3002\u672c\u6587\u5c06\u8be6\u7ec6\u4ecb\u7ecd […]","protected":false},"author":3,"featured_media":1032647,"comment_status":"closed","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[37],"tags":[],"acf":[],"_links":{"self":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/1032640"}],"collection":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/users\/3"}],"replies":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/comments?post=1032640"}],"version-history":[{"count":"1","href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/1032640\/revisions"}],"predecessor-version":[{"id":1032651,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/posts\/1032640\/revisions\/1032651"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media\/1032647"}],"wp:attachment":[{"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/media?parent=1032640"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/categories?post=1032640"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/docs.pingcode.com\/wp-json\/wp\/v2\/tags?post=1032640"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
![\"\u5982\u4f55\u5728python\u4e2d\u4f7f\u7528\u795e\u7ecf\u7f51\u7edc\"](\"https:\/\/cdn-docs.pingcode.com\/wp-content\/uploads\/2024\/12\/4331472e-1712-4958-9a0e-fed4d76d2b46.webp?x-oss-process=image\/auto-orient,1\/format,webp\")
<\/p>\n