test

.gitignore

@@ -1,3 +1,5 @@
 __pycache__
 main.py
 病马数据集*
+horse-colic.data
+horse-colic.test

function.py

@@ -1,5 +1,6 @@
 # function.py
 import numpy as np
+from sklearn.preprocessing import StandardScaler
 
 """
 挑战说明:
@@ -23,12 +24,12 @@ import numpy as np
 姓名请直接**填写中文**
 例如: name = "张三"
 """
-id = "202210423xxx"
-name = "test_name"
+id = "202210423229"
+name = "向兴宇"
 
 
 # 这里写自己的代码实现
-def train_svm(X, y, other_params="你自己设置一些其它参数"):
+def train_svm(X, y, C=0.20, max_iter=1000, learning_rate=0.01):
     """
     支持向量机训练函数。
 
@@ -40,9 +41,20 @@ def train_svm(X, y, other_params="你自己设置一些其它参数"):
     Returns:
     - numpy array: 训练得到的权重。
     """
+    m, n = X.shape
+    w = np.zeros(n)
+    b = 0
+    for i in range(max_iter):
+        for j in range(m):
+            if y[j] * (np.dot(X[j], w) + b) < 1:
+                w = w - learning_rate * (2 * w - C * y[j] * X[j])
+                b = b + learning_rate * C * y[j]
+            else:
+                w = w - learning_rate * 2 * w
+    return (w, b)
 
 
-def predict_svm(inX, weights):
+def predict_svm(inX, weight):
     """
     支持向量机预测函数。
 
@@ -53,6 +65,8 @@ def predict_svm(inX, weights):
     Returns:
     - int: 预测的分类标签。
     """
+    result = np.dot(inX, weight[0]) + weight[1]
+    return 1 if result > 0 else -1
 
 
 # 这里也写自己的代码实现
@@ -67,6 +81,22 @@ def load_data(filepath):
     - np.array: 特征数据。
     - np.array: 标签数据。
     """
+    # 使用 np.genfromtxt 加载数据，处理缺失值 '?' 为 -1
+    data = np.genfromtxt(filepath, delimiter=" ", missing_values="?", filling_values=0)
+
+    # 分离特征标签
+    labels = data[:, 22].astype(int)
+    # 删除无用特征
+    data = np.delete(data, [2] + list(range(22, 28)), axis=1)
+
+    # 合并死亡
+    labels = np.where((labels == 2) | (labels == 3), 0, labels)
+
+    # 数据标准化
+    scaler = StandardScaler()
+    data = scaler.fit_transform(data)
+
+    return data, labels
 
 
 if __name__ == "__main__":
@@ -77,9 +107,23 @@ if __name__ == "__main__":
     # 加载测试集
     X_test, y_test = load_data("./data/horse-colic.test")
 
-    # 训练支持向量机模型
-    weights = train_svm(X_train, y_train, learning_rate=0.01, iterations=1000)
+    before_acc = 0.0
+    for x in range(1, 2):
+        # 训练对率回归模型
+        np.random.seed(x)
+        weights = train_svm(X_train, y_train)
 
-    # 测试单个数据
-    y_pred = predict_svm(X_test[11], weights)
-    print(f"预测结果: {y_pred}, 实际结果: {y_test[11]}")
+        # 评估模型准确率
+        correct = 0
+        for i in range(X_test.shape[0]):
+            pred = predict_svm(X_test[i], weights)
+            if pred == y_test[i]:
+                correct += 1
+            # else:
+            #     print(f"{i+1}: P: {pred} A: {y_test[i]}")
+
+        # 准确率
+        acc = correct / X_test.shape[0]
+        if acc >= before_acc:
+            before_acc = acc
+            print(f"seed: {x} 准确率：{acc}")

requirements.txt

@@ -1,2 +1,3 @@
 numpy==2.1.1
 requests==2.32.3
+scikit-learn