2.3 · 机器学习基础与网络药理初步结合

数据（Data）+ 模型（Model）+ 评价（Evaluation）

from rdkit import Chem
from rdkit.Chem import Descriptors, Draw

# 从 SMILES 创建分子
mol = Chem.MolFromSmiles("CC(=O)Oc1ccccc1C(=O)O")  # 阿司匹林

# 计算常用描述符
mw = Descriptors.MolWt(mol)          # 分子量
logp = Descriptors.MolLogP(mol)       # 脂水分配系数
hbd = Descriptors.NumHDonors(mol)     # 氢键供体数
hba = Descriptors.NumHAcceptors(mol)  # 氢键受体数
print(f"分子量={mw:.1f}, LogP={logp:.1f}, HBD={hbd}, HBA={hba}")

from sklearn.preprocessing import StandardScaler

scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)

from sklearn.model_selection import train_test_split

X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42, stratify=y
)

数据分成 5 份：
[1][2][3][4][5]
第1轮：训练[2][3][4][5]  验证[1]
第2轮：训练[1][3][4][5]  验证[2]
...
第5轮：训练[1][2][3][4]  验证[5]
最终：5 次验证结果取平均

from sklearn.model_selection import cross_val_score
from sklearn.ensemble import RandomForestClassifier

model = RandomForestClassifier(random_state=42)
scores = cross_val_score(model, X, y, cv=5)
print(f"5折交叉验证平均准确率: {scores.mean():.3f}")

from sklearn.ensemble import RandomForestClassifier

rf = RandomForestClassifier(n_estimators=100, random_state=42)
rf.fit(X_train, y_train)

from sklearn.svm import SVC

svm = SVC(kernel="rbf", probability=True, random_state=42)
svm.fit(X_train, y_train)

from sklearn.neighbors import KNeighborsClassifier

knn = KNeighborsClassifier(n_neighbors=5)
knn.fit(X_train, y_train)

from sklearn.metrics import accuracy_score, precision_score, recall_score, f1_score

y_pred = rf.predict(X_test)
print("准确率:", accuracy_score(y_test, y_pred))
print("精确率:", precision_score(y_test, y_pred))
print("召回率:", recall_score(y_test, y_pred))
print("F1:", f1_score(y_test, y_pred))

from sklearn.metrics import roc_curve, auc
import matplotlib.pyplot as plt

# 预测概率
y_prob = rf.predict_proba(X_test)[:, 1]

# 计算 ROC
fpr, tpr, _ = roc_curve(y_test, y_prob)
roc_auc = auc(fpr, tpr)

# 画 ROC 曲线
plt.figure(figsize=(6, 5))
plt.plot(fpr, tpr, label=f"Random Forest (AUC = {roc_auc:.3f})")
plt.plot([0, 1], [0, 1], "k--", label="Random (AUC = 0.5)")
plt.xlabel("False Positive Rate")
plt.ylabel("True Positive Rate")
plt.title("ROC Curve")
plt.legend()
plt.savefig("roc_curve.png", dpi=300)
plt.show()

import pandas as pd

# 特征重要性
importance = rf.feature_importances_
feat_names = X.columns if hasattr(X, "columns") else [f"f{i}" for i in range(X.shape[1])]

# 排序并画图
imp_df = pd.DataFrame({"feature": feat_names, "importance": importance})
imp_df = imp_df.sort_values("importance", ascending=False)

plt.figure(figsize=(8, 5))
plt.barh(imp_df["feature"][:10], imp_df["importance"][:10])
plt.xlabel("Feature Importance")
plt.title("Top 10 Important Features")
plt.gca().invert_yaxis()
plt.savefig("feature_importance.png", dpi=300)
plt.show()

import numpy as np
import pandas as pd
from sklearn.model_selection import train_test_split, cross_val_score
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score, roc_curve, auc
import matplotlib.pyplot as plt

# ========== 1. 模拟数据（真实项目替换为 RDKit 计算的描述符表） ==========
rng = np.random.default_rng(42)
n = 200
X = pd.DataFrame({
    "MolWt": rng.uniform(150, 500, n),      # 分子量
    "LogP": rng.uniform(-2, 6, n),          # 脂水分配系数
    "HBD": rng.integers(0, 6, n),           # 氢键供体
    "HBA": rng.integers(0, 12, n),          # 氢键受体
    "TPSA": rng.uniform(20, 120, n),        # 极性表面积
})
y = ((X["MolWt"] > 250) & (X["LogP"] > 1) & (X["HBD"] <= 3)).astype(int)

# ========== 2. 划分训练/测试集 ==========
X_train, X_test, y_train, y_test = train_test_split(
    X, y, test_size=0.2, random_state=42, stratify=y
)

# ========== 3. 训练随机森林 ==========
rf = RandomForestClassifier(n_estimators=100, random_state=42)
rf.fit(X_train, y_train)

# ========== 4. 交叉验证 ==========
scores = cross_val_score(rf, X, y, cv=5)
print(f"5折交叉验证准确率: {scores.mean():.3f} ± {scores.std():.3f}")

# ========== 5. 测试集评价 ==========
y_pred = rf.predict(X_test)
print(f"测试集准确率: {accuracy_score(y_test, y_pred):.3f}")

# ========== 6. ROC 曲线 ==========
y_prob = rf.predict_proba(X_test)[:, 1]
fpr, tpr, _ = roc_curve(y_test, y_prob)
roc_auc = auc(fpr, tpr)

plt.figure(figsize=(6, 5))
plt.plot(fpr, tpr, label=f"Random Forest (AUC = {roc_auc:.3f})")
plt.plot([0, 1], [0, 1], "k--")
plt.xlabel("False Positive Rate")
plt.ylabel("True Positive Rate")
plt.title("ROC Curve")
plt.legend()
plt.savefig("roc_curve.png", dpi=300)

# ========== 7. 特征重要性 ==========
imp = pd.Series(rf.feature_importances_, index=X.columns).sort_values(ascending=False)
print("特征重要性:\n", imp)

# 划分
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42, stratify=y)

# 交叉验证
from sklearn.model_selection import cross_val_score
scores = cross_val_score(model, X, y, cv=5)

# 随机森林
from sklearn.ensemble import RandomForestClassifier
rf = RandomForestClassifier(n_estimators=100, random_state=42).fit(X_train, y_train)

# 评价
from sklearn.metrics import accuracy_score, roc_curve, auc