![\"\"](\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2025\/08\/20250810051420-68982aac54013.png\")
![\"\"](\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2025\/08\/20250810051420-68982aacb7f65.png\")
<\/p>\n<\/p>\n\u6709\u76d1\u7763\u5b66\u4e60\u6709\u5916\u90e8\u7ed3\u679cy，\u65e0\u76d1\u7763\u5b66\u4e60\u65e0\u5916\u90e8\u7ed3\u679cy，\u6b64\u4e3a\u4e8c\u8005\u6838\u5fc3\u533a\u522b\u3002<\/p>\n
\u6982\u5ff5：
\n\u57fa\u4e8e\u5bc6\u5ea6\u7684\u5e26\u566a\u58f0\u7684\u7a7a\u95f4\u805a\u7c7b\u5e94\u7528\u7b97\u6cd5，\u5b83\u662f\u5c06\u7c07\u5b9a\u4e49\u4e3a\u5bc6\u5ea6\u76f8\u8fde\u7684\u70b9\u7684\u6700\u5927\u96c6\u5408，\u80fd\u591f\u628a\u5177\u6709\u8db3\u591f\u9ad8\u5bc6\u5ea6\u7684\u533a\u57df\u5212\u5206\u4e3a\u7c07，\u5e76\u5728\u566a\u58f0\u7684\u7a7a\u95f4\u6570\u636e\u96c6\u4e2d\u53d1\u73b0\u4efb\u610f\u5f62\u72b6\u7684\u805a\u7c7b\u3002<\/p>\n<\/p>\n
<\/p>\n<\/p>\n
\u8981\u70b9：
\n1. \u6838\u5fc3\u5bf9\u8c61：A \u70b9
\n2.E \u90bb\u57df：\u7ed9\u5b9a\u5bf9\u8c61\u534a\u5f84\u4e3a E \u5185\u7684\u533a\u57df
\n3. \u76f4\u63a5\u5bc6\u5ea6\u53ef\u8fbe：
\n4. \u5bc6\u5ea6\u53ef\u8fbe：
\n5. \u8fb9\u754c\u70b9：B \u70b9\u3001C \u70b9
\n6. \u79bb\u7fa4\u70b9：N \u70b9<\/p>\n
![\"\"](\"https:\/\/www.wsisp.com\/helps\/wp-content\/uploads\/2025\/08\/20250810051421-68982aad20529.png\")
<\/p>\n
python<\/p>\n
class sklearn.cluster.DBSCAN(
\n eps=0.5,
\n min_samples=5,
\n metric='euclidean',
\n metric_params=None,
\n algorithm='auto',
\n leaf_size=30,
\n p=None,
\n n_jobs=None
\n)<\/p>\n
import pandas as pd
\nfrom sklearn.cluster import DBSCAN
\nfrom sklearn import metrics<\/p>\n
# \u8bfb\u53d6\u6587\u4ef6
\nbeer = pd.read_table("data.txt", sep=' ', encoding='utf8', engine='python')
\n# \u4f20\u5165\u53d8\u91cf（\u5217\u540d）
\nX = beer[["calories", "sodium", "alcohol", "cost"]]
\n# DBSCAN\u805a\u7c7b\u5206\u6790
\n"""
\neps:\u534a\u5f84
\nmin_samples: \u6700\u5c0f\u5bc6\u5ea6 \u3010\u5c31\u662f\u5706\u5185\u6700\u5c11\u6709\u51e0\u4e2a\u6837\u672c\u70b9\u3011
\nlabels: \u5206\u7c7b\u7ed3\u679c \u3010\u81ea\u52a8\u5206\u7c7b，-1\u4e3a\u79bb\u7fa4\u70b9\u3011
\n"""
\ndb = DBSCAN(eps=20, min_samples=2).fit(X)#\u5f52\u4e00\u5316，
\nlabels = db.labels_<\/p>\n
# \u6dfb\u52a0\u7ed3\u679c\u81f3\u539f\u6570\u636e\u6846
\n"""
\nmetrics.silhouette_score\u8f6e\u5ed3\u8bc4\u4ef7\u51fd\u6570，\u5b83\u662f\u805a\u7c7b\u6a21\u578b\u4f18\u52a3\u7684\u4e00\u79cd\u8bc4\u4f30\u65b9\u5f0f，\u53ef\u7528\u4e8e\u5bf9\u805a\u7c7b\u7ed3\u679c\u8fdb\u884c\u8bc4
\nX:\u6570\u636e\u96c6 scaled_cluster: \u805a\u7c7b\u7ed3\u679c
\nscore: \u975e\u6807\u51c6\u5316\u805a\u7c7b\u7ed3\u679c\u7684\u8f6e\u5ed3\u7cfb\u6570->\u805a\u7c7b
\n"""score = metrics.silhouette_score(X, beer.cluster_db)
\nprint(score)<\/p>\n
#\u4f5c\u4e1a: KNN\u7684\u5bdd\u5ba4\u3002 y \u6682\u65f6\u4e0d\u7528y，\u7528kmens\u6216dbscan\u6765\u8bad\u7ec3\u3002 \u5c06\u4e00\u90e8\u5206\u4f5c\u4e3a\u6d4b\u8bd5\u96c6，\u6d4b\u8bd5\u4e00\u4e0b\u6548\u679c<\/p>\n
\u9996\u5148\u5bfc\u5165\u4e86\u8fdb\u884c\u6570\u636e\u5206\u6790\u548c\u805a\u7c7b\u6240\u9700\u7684\u5e93：<\/p>\n<\/p>\n
python<\/p>\n
\u8fd0\u884c<\/p>\n
import pandas as pd
\nfrom sklearn.cluster import DBSCAN
\nfrom sklearn import metrics<\/p>\n
\u4f7f\u7528 pandas \u7684read_table\u51fd\u6570\u8bfb\u53d6\u6570\u636e\u6587\u4ef6：<\/p>\n<\/p>\n
python<\/p>\n
\u8fd0\u884c<\/p>\n
# \u8bfb\u53d6\u6587\u4ef6
\nbeer = pd.read_table("data.txt", sep=' ', encoding='utf8', engine='python')<\/p>\n
\u4ece\u8bfb\u53d6\u7684\u6570\u636e\u4e2d\u9009\u53d6\u9700\u8981\u7528\u4e8e\u805a\u7c7b\u5206\u6790\u7684\u7279\u5f81\u5217：<\/p>\n<\/p>\n
python<\/p>\n
\u8fd0\u884c<\/p>\n
# \u4f20\u5165\u53d8\u91cf（\u5217\u540d）
\nX = beer[["calories", "sodium", "alcohol", "cost"]]<\/p>\n
\u4f7f\u7528 DBSCAN \u7b97\u6cd5\u8fdb\u884c\u805a\u7c7b\u5206\u6790：<\/p>\n<\/p>\n
python<\/p>\n
\u8fd0\u884c<\/p>\n
# DBSCAN\u805a\u7c7b\u5206\u6790
\n"""
\neps:\u534a\u5f84
\nmin_samples: \u6700\u5c0f\u5bc6\u5ea6 \u3010\u5c31\u662f\u5706\u5185\u6700\u5c11\u6709\u51e0\u4e2a\u6837\u672c\u70b9\u3011
\nlabels: \u5206\u7c7b\u7ed3\u679c \u3010\u81ea\u52a8\u5206\u7c7b，-1\u4e3a\u79bb\u7fa4\u70b9\u3011
\n"""
\ndb = DBSCAN(eps=20, min_samples=2).fit(X)#\u5f52\u4e00\u5316，
\nlabels = db.labels_<\/p>\n
\u4f7f\u7528\u8f6e\u5ed3\u7cfb\u6570（silhouette score）\u8bc4\u4f30\u805a\u7c7b\u6548\u679c：<\/p>\n<\/p>\n
python<\/p>\n
\u8fd0\u884c<\/p>\n
# \u8ba1\u7b97\u8f6e\u5ed3\u7cfb\u6570\u8bc4\u4f30\u805a\u7c7b\u7ed3\u679c
\n"""
\nmetrics.silhouette_score\u8f6e\u5ed3\u8bc4\u4ef7\u51fd\u6570，\u5b83\u662f\u805a\u7c7b\u6a21\u578b\u4f18\u52a3\u7684\u4e00\u79cd\u8bc4\u4f30\u65b9\u5f0f，\u53ef\u7528\u4e8e\u5bf9\u805a\u7c7b\u7ed3\u679c\u8fdb\u884c\u8bc4
\nX:\u6570\u636e\u96c6 labels: \u805a\u7c7b\u7ed3\u679c
\nscore: \u805a\u7c7b\u7ed3\u679c\u7684\u8f6e\u5ed3\u7cfb\u6570
\n"""
\nscore = metrics.silhouette_score(X, labels)
\nprint(score)<\/p>\n
\u539f\u4ee3\u7801\u4e2d\u6709\u4e00\u5904\u7b14\u8bef，beer.cluster_db\u5e94\u8be5\u6539\u4e3alabels，\u56e0\u4e3a\u805a\u7c7b\u7ed3\u679c\u5b58\u50a8\u5728labels\u53d8\u91cf\u4e2d，\u800c\u4e0d\u662fbeer\u6570\u636e\u6846\u7684cluster_db\u5217\u4e2d（\u8be5\u5217\u5c1a\u672a\u521b\u5efa）\u3002\u4e0a\u9762\u7684\u4ee3\u7801\u5df2\u7ecf\u4fee\u6b63\u4e86\u8fd9\u4e00\u95ee\u9898\u3002<\/p>\n<\/p>\n
\u5982\u679c\u9700\u8981\u5c06\u805a\u7c7b\u7ed3\u679c\u6dfb\u52a0\u5230\u539f\u6570\u636e\u6846\u4e2d，\u53ef\u4ee5\u6dfb\u52a0\u4ee5\u4e0b\u4ee3\u7801：<\/p>\n<\/p>\n
python<\/p>\n
\u8fd0\u884c<\/p>\n
# \u6dfb\u52a0\u805a\u7c7b\u7ed3\u679c\u5230\u539f\u6570\u636e\u6846
\nbeer['cluster_db'] = labels<\/p>\n
\u4e0a\u9762\u5462\u662f\u8f6e\u5ed3\u7cfb\u6570，\u4e0b\u9762\u7684\u662f\u5206\u7c7b\u6307\u6807\u505a\u7684<\/p>\n
\u603b\u4ee3\u7801<\/p>\n
import os
\nimport pandas as pd<\/p>\n
file_path = "datingTestSet2.txt"
\nif not os.path.exists(file_path):
\n print(f"Error: File '{file_path}' not found")
\n exit(1)<\/p>\n
try:
\n data = pd.read_csv(file_path, sep=None, engine='python')
\nexcept Exception as e:
\n print(f"File read error: {e}")
\n exit(1)<\/p>\n
X = data.iloc[:, :-1]
\ny_true = data.iloc[:, -1].astype(int)<\/p>\n
import numpy as np
\nfrom scipy.optimize import linear_sum_assignment<\/p>\n
def align_labels(true_labels, cluster_labels):
\n unique_true = np.unique(true_labels)
\n unique_cluster = np.unique(cluster_labels)
\n n_true = len(unique_true)
\n n_cluster = len(unique_cluster)<\/p>\n
# \u53d6\u8f83\u5c0f\u7684\u7c7b\u522b\u6570\u4f5c\u4e3a\u77e9\u9635\u7ef4\u5ea6，\u907f\u514d\u4e0d\u5339\u914d
\n n = min(n_true, n_cluster)
\n cost_matrix = np.zeros((n, n))<\/p>\n
for i in range(n):
\n for j in range(n):
\n cost_matrix[i, j] = -np.sum((true_labels == unique_true[i]) & (cluster_labels == unique_cluster[j]))<\/p>\n
row_ind, col_ind = linear_sum_assignment(cost_matrix)
\n label_map = {}
\n # \u53ea\u6620\u5c04\u5b58\u5728\u7684\u7d22\u5f15，\u907f\u514d\u8d8a\u754c
\n for i, j in zip(row_ind, col_ind):
\n if j < len(unique_cluster) and i < len(unique_true):
\n label_map[unique_cluster[j]] = unique_true[i]<\/p>\n
aligned_labels = np.array([label_map.get(c, -1) for c in cluster_labels])
\n return aligned_labels<\/p>\n
from sklearn.cluster import KMeans
\nimport matplotlib.pyplot as plt
\nfrom sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score<\/p>\n
k_values = [2, 3, 4, 5]
\nmetrics_kmeans = {'accuracy': [], 'precision': [], 'recall': [], 'f1': []}<\/p>\n
for k in k_values:
\n kmeans = KMeans(n_clusters=k, random_state=42).fit(X)
\n aligned_labels = align_labels(y_true, kmeans.labels_)<\/p>\n
mask = aligned_labels != -1
\n if np.sum(mask) > 0:
\n acc = accuracy_score(y_true[mask], aligned_labels[mask])
\n prec = precision_score(y_true[mask], aligned_labels[mask], average='macro', zero_division=0)
\n rec = recall_score(y_true[mask], aligned_labels[mask], average='macro', zero_division=0)
\n f1 = f1_score(y_true[mask], aligned_labels[mask], average='macro', zero_division=0)<\/p>\n
metrics_kmeans['accuracy'].append(acc)
\n metrics_kmeans['precision'].append(prec)
\n metrics_kmeans['recall'].append(rec)
\n metrics_kmeans['f1'].append(f1)<\/p>\n
print(f"KMeans(k={k}) – Acc: {acc:.3f}, Prec: {prec:.3f}, Rec: {rec:.3f}, F1: {f1:.3f}")<\/p>\n
plt.figure(figsize=(12, 8))
\nfor i, metric in enumerate(metrics_kmeans.keys()):
\n plt.subplot(2, 2, i + 1)
\n plt.plot(k_values, metrics_kmeans[metric], 'o-')
\n plt.xlabel('k')
\n plt.ylabel(metric)
\n plt.title(f'KMeans {metric}')<\/p>\n
plt.tight_layout()
\nplt.show()<\/p>\n
from sklearn.cluster import DBSCAN<\/p>\n
db = DBSCAN(eps=0.5, min_samples=5).fit(X)
\ndb_labels = db.labels_<\/p>\n
mask_noise = db_labels != -1
\nif np.sum(mask_noise) > 0:
\n aligned_db_labels = align_labels(y_true[mask_noise], db_labels[mask_noise])<\/p>\n
acc_db = accuracy_score(y_true[mask_noise], aligned_db_labels)
\n prec_db = precision_score(y_true[mask_noise], aligned_db_labels, average='macro', zero_division=0)
\n rec_db = recall_score(y_true[mask_noise], aligned_db_labels, average='macro', zero_division=0)
\n f1_db = f1_score(y_true[mask_noise], aligned_db_labels, average='macro', zero_division=0)<\/p>\n
print(f"\\\\nDBSCAN – Acc: {acc_db:.3f}, Prec: {prec_db:.3f}, Rec: {rec_db:.3f}, F1: {f1_db:.3f}")
\nelse:
\n print("\\\\nDBSCAN: Too many noise points, cannot calculate metrics")<\/p>\n
data['db_cluster'] = db_labels
\nprint("\\\\nSample results:")
\nprint(data.head(10))<\/p>\n
\u8fd9\u6bb5\u4ee3\u7801\u662f\u4e0a\u4e00\u4e2a\u805a\u7c7b\u5206\u6790\u4ee3\u7801\u7684\u6539\u8fdb\u7248\u672c，\u6838\u5fc3\u533a\u522b\u5728\u4e8e\u8bc4\u4f30\u6307\u6807\u7684\u53d8\u5316：\u4e0a\u4e00\u4e2a\u4ee3\u7801\u4f7f\u7528\u8f6e\u5ed3\u7cfb\u6570（\u805a\u7c7b\u4e13\u7528\u6307\u6807，\u9002\u7528\u4e8e\u65e0\u771f\u5b9e\u6807\u7b7e\u7684\u60c5\u51b5），\u800c\u672c\u4ee3\u7801\u56e0\u4e3a\u6709\u771f\u5b9e\u6807\u7b7e（y_true），\u6240\u4ee5\u91c7\u7528\u4e86\u5206\u7c7b\u4efb\u52a1\u4e2d\u5e38\u7528\u7684\u8bc4\u4f30\u6307\u6807（\u51c6\u786e\u7387\u3001\u7cbe\u786e\u7387\u3001\u53ec\u56de\u7387\u3001F1 \u503c）\u6765\u66f4\u76f4\u89c2\u5730\u8861\u91cf\u805a\u7c7b\u6548\u679c\u4e0e\u771f\u5b9e\u7c7b\u522b\u7684\u5339\u914d\u7a0b\u5ea6\u3002<\/p>\n
\u9996\u5148\u5bfc\u5165\u57fa\u7840\u5de5\u5177\u5e93，\u68c0\u67e5\u5e76\u8bfb\u53d6\u6570\u636e\u6587\u4ef6：<\/p>\n<\/p>\n
python<\/p>\n
\u8fd0\u884c<\/p>\n
import os
\nimport pandas as pd<\/p>\n
file_path = "datingTestSet2.txt"
\nif not os.path.exists(file_path):
\n print(f"Error: File '{file_path}' not found")
\n exit(1)<\/p>\n
try:
\n data = pd.read_csv(file_path, sep=None, engine='python') # \u81ea\u52a8\u8bc6\u522b\u5206\u9694\u7b26\u8bfb\u53d6\u6570\u636e
\nexcept Exception as e:
\n print(f"File read error: {e}")
\n exit(1)<\/p>\n
\u4ece\u8bfb\u53d6\u7684\u6570\u636e\u4e2d\u62c6\u5206\u51fa\u7279\u5f81\u548c\u771f\u5b9e\u6807\u7b7e（\u7528\u4e8e\u540e\u7eed\u8bc4\u4f30\u805a\u7c7b\u6548\u679c）：<\/p>\n<\/p>\n
python<\/p>\n
\u8fd0\u884c<\/p>\n
X = data.iloc[:, :-1] # \u7279\u5f81\u6570\u636e（\u6240\u6709\u7279\u5f81\u5217）
\ny_true = data.iloc[:, -1].astype(int) # \u771f\u5b9e\u6807\u7b7e（\u6700\u540e\u4e00\u5217）<\/p>\n
\u805a\u7c7b\u7b97\u6cd5\u8f93\u51fa\u7684\u6807\u7b7e\u4ec5\u4ee3\u8868\u7c7b\u522b\u7f16\u53f7（\u4e0e\u771f\u5b9e\u6807\u7b7e\u7684\u7f16\u53f7\u53ef\u80fd\u4e0d\u4e00\u81f4），\u9700\u8981\u901a\u8fc7\u6807\u7b7e\u5bf9\u9f50\u5c06\u805a\u7c7b\u6807\u7b7e\u6620\u5c04\u5230\u771f\u5b9e\u6807\u7b7e：<\/p>\n<\/p>\n
python<\/p>\n
\u8fd0\u884c<\/p>\n
import numpy as np
\nfrom scipy.optimize import linear_sum_assignment<\/p>\n
align_labels\u51fd\u6570\u7684\u4f5c\u7528\u662f\u5c06\u805a\u7c7b\u8f93\u51fa\u7684\u6807\u7b7e（cluster_labels）\u4e0e\u771f\u5b9e\u6807\u7b7e（true_labels）\u8fdb\u884c\u6620\u5c04，\u89e3\u51b3 \u201c\u7c7b\u522b\u7f16\u53f7\u4e0d\u5339\u914d\u201d \u95ee\u9898：<\/p>\n<\/p>\n
python<\/p>\n
\u8fd0\u884c<\/p>\n
def align_labels(true_labels, cluster_labels):
\n unique_true = np.unique(true_labels) # \u771f\u5b9e\u6807\u7b7e\u7684\u552f\u4e00\u503c
\n unique_cluster = np.unique(cluster_labels) # \u805a\u7c7b\u6807\u7b7e\u7684\u552f\u4e00\u503c
\n n_true = len(unique_true)
\n n_cluster = len(unique_cluster)<\/p>\n
# \u53d6\u8f83\u5c0f\u7684\u7c7b\u522b\u6570\u4f5c\u4e3a\u77e9\u9635\u7ef4\u5ea6，\u907f\u514d\u4e0d\u5339\u914d
\n n = min(n_true, n_cluster)
\n cost_matrix = np.zeros((n, n))<\/p>\n
# \u6784\u5efa\u6210\u672c\u77e9\u9635（\u503c\u8d8a\u5c0f\u8868\u793a\u5339\u914d\u5ea6\u8d8a\u9ad8）
\n for i in range(n):
\n for j in range(n):
\n # \u8ba1\u7b97\u805a\u7c7b\u6807\u7b7ej\u4e0e\u771f\u5b9e\u6807\u7b7ei\u7684\u91cd\u53e0\u6570\u91cf（\u53d6\u8d1f\u53f7\u8f6c\u4e3a\u6210\u672c）
\n cost_matrix[i, j] = -np.sum((true_labels == unique_true[i]) & (cluster_labels == unique_cluster[j]))<\/p>\n
# \u7528\u5308\u7259\u5229\u7b97\u6cd5\u627e\u5230\u6700\u4f18\u6620\u5c04
\n row_ind, col_ind = linear_sum_assignment(cost_matrix)
\n label_map = {}
\n # \u6784\u5efa\u6620\u5c04\u5173\u7cfb（\u805a\u7c7b\u6807\u7b7e\u2192\u771f\u5b9e\u6807\u7b7e）
\n for i, j in zip(row_ind, col_ind):
\n if j < len(unique_cluster) and i < len(unique_true):
\n label_map[unique_cluster[j]] = unique_true[i]<\/p>\n
# \u5c06\u805a\u7c7b\u6807\u7b7e\u8f6c\u6362\u4e3a\u6620\u5c04\u540e\u7684\u771f\u5b9e\u6807\u7b7e\u683c\u5f0f
\n aligned_labels = np.array([label_map.get(c, -1) for c in cluster_labels])
\n return aligned_labels<\/p>\n
\u5bfc\u5165 KMeans \u7b97\u6cd5\u3001\u7ed8\u56fe\u5de5\u5177\u548c\u5206\u7c7b\u8bc4\u4f30\u6307\u6807：<\/p>\n<\/p>\n
python<\/p>\n
\u8fd0\u884c<\/p>\n
from sklearn.cluster import KMeans
\nimport matplotlib.pyplot as plt
\nfrom sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score<\/p>\n
\u5c1d\u8bd5\u4e0d\u540c\u7684\u805a\u7c7b\u6570（k=2,3,4,5），\u5bf9\u6bcf\u4e2ak\u8fdb\u884c\u805a\u7c7b\u5e76\u8bc4\u4f30：<\/p>\n
python\u8fd0\u884c<\/p>\n
k_values = [2, 3, 4, 5] # \u5c1d\u8bd5\u7684\u805a\u7c7b\u6570
\nmetrics_kmeans = {'accuracy': [], 'precision': [], 'recall': [], 'f1': []} # \u5b58\u50a8\u5404\u6307\u6807\u7ed3\u679c<\/p>\n
for k in k_values:
\n # \u8bad\u7ec3KMeans\u6a21\u578b
\n kmeans = KMeans(n_clusters=k, random_state=42).fit(X)
\n # \u5bf9\u9f50\u805a\u7c7b\u6807\u7b7e\u4e0e\u771f\u5b9e\u6807\u7b7e
\n aligned_labels = align_labels(y_true, kmeans.labels_)<\/p>\n
# \u8fc7\u6ee4\u65e0\u6cd5\u6620\u5c04\u7684\u6807\u7b7e（\u4ec5\u4fdd\u7559\u6709\u6548\u6807\u7b7e）
\n mask = aligned_labels != -1
\n if np.sum(mask) > 0:
\n # \u8ba1\u7b97\u8bc4\u4f30\u6307\u6807
\n acc = accuracy_score(y_true[mask], aligned_labels[mask])
\n prec = precision_score(y_true[mask], aligned_labels[mask], average='macro', zero_division=0)
\n rec = recall_score(y_true[mask], aligned_labels[mask], average='macro', zero_division=0)
\n f1 = f1_score(y_true[mask], aligned_labels[mask], average='macro', zero_division=0)<\/p>\n
# \u5b58\u50a8\u6307\u6807
\n metrics_kmeans['accuracy'].append(acc)
\n metrics_kmeans['precision'].append(prec)
\n metrics_kmeans['recall'].append(rec)
\n metrics_kmeans['f1'].append(f1)<\/p>\n
# \u6253\u5370\u7ed3\u679c
\n print(f"KMeans(k={k}) – Acc: {acc:.3f}, Prec: {prec:.3f}, Rec: {rec:.3f}, F1: {f1:.3f}")<\/p>\n
\u7528\u5b50\u56fe\u5c55\u793a\u4e0d\u540ck\u503c\u4e0b\u5404\u8bc4\u4f30\u6307\u6807\u7684\u53d8\u5316\u8d8b\u52bf，\u76f4\u89c2\u5bf9\u6bd4\u805a\u7c7b\u6548\u679c：<\/p>\n
python\u8fd0\u884c<\/p>\n
plt.figure(figsize=(12, 8))
\nfor i, metric in enumerate(metrics_kmeans.keys()):
\n plt.subplot(2, 2, i + 1) # 2\u884c2\u5217\u5b50\u56fe
\n plt.plot(k_values, metrics_kmeans[metric], 'o-') # \u7ed8\u5236\u6298\u7ebf\u56fe
\n plt.xlabel('k') # x\u8f74：\u805a\u7c7b\u6570k
\n plt.ylabel(metric) # y\u8f74：\u8bc4\u4f30\u6307\u6807
\n plt.title(f'KMeans {metric}') # \u5b50\u56fe\u6807\u9898<\/p>\n
plt.tight_layout() # \u8c03\u6574\u5e03\u5c40
\nplt.show() # \u663e\u793a\u56fe\u50cf<\/p>\n
\u4f7f\u7528 DBSCAN \u7b97\u6cd5\u8fdb\u884c\u805a\u7c7b，\u540c\u6837\u901a\u8fc7\u6807\u7b7e\u5bf9\u9f50\u548c\u5206\u7c7b\u6307\u6807\u8bc4\u4f30：<\/p>\n<\/p>\n
from sklearn.cluster import DBSCAN<\/p>\n
# \u8bad\u7ec3DBSCAN\u6a21\u578b（eps：\u534a\u5f84，min_samples：\u6700\u5c0f\u6837\u672c\u6570）
\ndb = DBSCAN(eps=0.5, min_samples=5).fit(X)
\ndb_labels = db.labels_ # \u83b7\u53d6\u805a\u7c7b\u6807\u7b7e（-1\u4e3a\u566a\u58f0\u70b9）<\/p>\n
# \u8fc7\u6ee4\u566a\u58f0\u70b9
\nmask_noise = db_labels != -1
\nif np.sum(mask_noise) > 0:
\n # \u5bf9\u9f50\u6709\u6548\u6837\u672c\u7684\u6807\u7b7e
\n aligned_db_labels = align_labels(y_true[mask_noise], db_labels[mask_noise])<\/p>\n
# \u8ba1\u7b97\u8bc4\u4f30\u6307\u6807
\n acc_db = accuracy_score(y_true[mask_noise], aligned_db_labels)
\n prec_db = precision_score(y_true[mask_noise], aligned_db_labels, average='macro', zero_division=0)
\n rec_db = recall_score(y_true[mask_noise], aligned_db_labels, average='macro', zero_division=0)
\n f1_db = f1_score(y_true[mask_noise], aligned_db_labels, average='macro', zero_division=0)<\/p>\n
# \u6253\u5370\u7ed3\u679c
\n print(f"\\\\nDBSCAN – Acc: {acc_db:.3f}, Prec: {prec_db:.3f}, Rec: {rec_db:.3f}, F1: {f1_db:.3f}")
\nelse:
\n print("\\\\nDBSCAN: Too many noise points, cannot calculate metrics")<\/p>\n
\u5c06 DBSCAN \u7684\u805a\u7c7b\u7ed3\u679c\u6dfb\u52a0\u5230\u539f\u59cb\u6570\u636e\u4e2d，\u6253\u5370\u524d 10 \u884c\u6837\u672c，\u76f4\u89c2\u67e5\u770b\u805a\u7c7b\u6807\u7b7e：python<\/p>\n
\u8fd0\u884c<\/p>\n
data['db_cluster'] = db_labels # \u6dfb\u52a0DBSCAN\u805a\u7c7b\u7ed3\u679c\u5230\u6570\u636e\u6846
\nprint("\\\\nSample results:")
\nprint(data.head(10)) # \u663e\u793a\u524d10\u884c\u6570\u636e\u53ca\u805a\u7c7b\u6807\u7b7e<\/p>\n","protected":false},"excerpt":{"rendered":"
\u6587\u7ae0\u6d4f\u89c8\u9605\u8bfb328\u6b21\uff0c\u70b9\u8d5e3\u6b21\uff0c\u6536\u85cf3\u6b21\u3002\u672c\u6587\u4ecb\u7ecd\u4e86DBSCAN\u5bc6\u5ea6\u805a\u7c7b\u7b97\u6cd5\u7684\u539f\u7406\u3001\u53c2\u6570\u8c03\u4f18\u53ca\u5b9e\u73b0\u8fc7\u7a0b\u3002DBSCAN\u901a\u8fc7\u5b9a\u4e49\u6838\u5fc3\u5bf9\u8c61\u3001E\u90bb\u57df\u548c\u5bc6\u5ea6\u53ef\u8fbe\u7b49\u6982\u5ff5\uff0c\u80fd\u591f\u53d1\u73b0\u4efb\u610f\u5f62\u72b6\u7684\u805a\u7c7b\u5e76\u8bc6\u522b\u566a\u58f0\u70b9\u3002\u6587\u4e2d\u8be6\u7ec6\u8bf4\u660e\u4e86Python\u4e2dsklearn.cluster.DBSCAN\u7684\u53c2\u6570\u8bbe\u7f6e\uff0c\u5305\u62eceps\u90bb\u57df\u534a\u5f84\u3001min_samples\u6700\u5c0f\u6837\u672c\u6570\u7b49\u5173\u952e\u53c2\u6570\u7684\u5f71\u54cd\u3002\u540c\u65f6\u63d0\u4f9b\u4e86\u5b8c\u6574\u7684\u4ee3\u7801\u5b9e\u73b0\uff0c\u5305\u62ec\u6570\u636e\u9884\u5904\u7406\u3001DBSCAN\u6a21\u578b\u8bad\u7ec3\u3001\u8f6e\u5ed3\u7cfb\u6570\u8bc4\u4f30\u7b49\u6b65\u9aa4\uff0c\u5e76\u5bf9\u6bd4\u4e86KMeans\u4e0eDBSCAN\u5728\u771f\u5b9e\u6570\u636e\u96c6\u4e0a\u7684\u8868\u73b0\u3002\u6700\u540e\u901a\u8fc7\u6807\u7b7e\u5bf9\u9f50\u65b9\u6cd5\u89e3\u51b3\u4e86\u805a\u7c7b\u6807\u7b7e\u4e0e\u771f\u5b9e\u6807\u7b7e\u7684\u6620\u5c04\u95ee\u9898\uff0c\u91c7\u7528\u5206\u7c7b\u6307<\/p>\n","protected":false},"author":2,"featured_media":50860,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[50,207,427],"topic":[],"class_list":["post-50863","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-server","tag-50","tag-207","tag-427"],"yoast_head":"\n