从 ROC 曲线评估准确率

1# 使用 NetworkX 创建有向图
2G = nx.DiGraph()
3
4# 添加节点和边
5G.add_nodes_from(nodes + maturity_nodes)
6G.add_edges_from(Edges + maturity_edges)
7
8# 定义节点的笛卡尔坐标
9pos = {
10    "A": (0, 2), "B": (1, 2), "C": (0, 1),
11    "D": (2, 2), "E": (2, 0), "F": (0, 0),
12    "G": (3, 1), "H": (4, 0), "I": (2, 1),
13    "J": (1, 1), "Project_Overcost": (2, -1)
14}
15
16# 定义成熟度节点的位置
17maturity_pos = {f"Maturity_{node}": (pos[node][0] -0.45, pos[node][1] + 0.15) for node in nodes}
18pos.update(maturity_pos)
19
20# 绘制图形
21plt.figure(figsize=(12, 8))
22nx.draw(G, pos, with_labels=True, node_size=2000, node_color="lightblue", font_size=12, font_weight='bold')
23
24plt.show()
25plt.savefig('output/FinalModel')

1# 将数据拆分为训练集和测试集
2train_df, test_df = train_test_split(df, test_size=0.15, random_state=12)
3
4# 使用离散化数据拟合贝叶斯网络
5bn = BayesianNetwork(sm)
6bn = bn.fit_node_states(train_df)
7
8# 使用 BayesianEstimator 拟合 CPDs
9bn = bn.fit_cpds(train_df, method="BayesianEstimator", bayes_prior="K2")
10#bn = bn.fit_cpds(train_df, method="BayesianEstimator", bayes_prior="BDeu",equivalent_sample_size=5)
11
12# 创建一个使用 InferenceEngine 进行预测的函数：
13ie = InferenceEngine(bn)
14
15def make_predictions(test_df):
16    predictions = []
17
18    for _, row in test_df.iterrows():
19        query = {col: row[col] for col in train_df.columns if col != 'Project_Overcost'}
20        marginals = ie.query(query)
21        pred_prob = marginals["Project_Overcost"][1]
22        predictions.append(pred_prob)
23
24    return np.array(predictions)
25
26# 进行预测并计算 ROC 曲线
27y_true = test_df['Project_Overcost'].values
28y_pred = make_predictions(test_df)
29fpr, tpr, _ = roc_curve(y_true, y_pred)
30roc_auc = auc(fpr, tpr)