更新 function.py

function.py

@@ -1,5 +1,6 @@
 # function.py
 import numpy as np
+from sklearn.model_selection import train_test_split
 
 """
 实验说明:
@@ -67,11 +68,10 @@ def train_logreg(X, y, learning_rate=0.001, iterations=1000, batch_size=32, init
     """
 
 
-def predict_logreg(inX, weights):
+def evaluate_model(X_val, y_val, weights):
-    inX = np.insert(inX, 0, 1)  # 添加偏置项
+    predictions = np.array([predict_logreg(x, weights) for x in X_val])
-    z = np.dot(inX, weights)
+    accuracy = np.mean(predictions == y_val)
-    h = sigmoid(z)
+    return accuracy
-    return 1 if h >= 0.5 else 0
 
     """
     对率回归预测函数。
@@ -118,30 +118,40 @@ def load_data(filepath):
 
 
 if __name__ == "__main__":
-    # 加载训练集
+    # 加载数据并分割为训练集和验证集
-    X_train, y_train = load_data("./data/horse-colic.data")
+    X, y = load_data("./data/horse-colic.data")
-
+    X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.2, random_state=42)
-    # 加载测试集
-    X_test, y_test = load_data("./data/horse-colic.test")
 
     # 模型调优参数
     learning_rates = [0.01, 0.001, 0.0001]
     iterations = [500, 1000, 2000]
     batch_sizes = [16, 32, 64]
-    initial_weights = np.random.randn(X_train.shape[1] + 1)  # 随机初始化权重
+    
+    best_accuracy = 0
+    best_params = {}
 
-    # 选择一组参数进行训练
+    # 网格搜索
-    chosen_learning_rate = learning_rates[1]
+    for lr in learning_rates:
-    chosen_iterations = iterations[1]
+        for it in iterations:
-    chosen_batch_size = batch_sizes[1]
+            for bs in batch_sizes:
+                initial_weights = np.random.randn(X_train.shape[1] + 1)
+                weights = train_logreg(X_train, y_train, lr, it, bs, initial_weights)
+                accuracy = evaluate_model(X_val, y_val, weights)
+                
+                print(f"Learning Rate: {lr}, Iterations: {it}, Batch Size: {bs}, Accuracy: {accuracy}")
+                
+                if accuracy > best_accuracy:
+                    best_accuracy = accuracy
+                    best_params = {'learning_rate': lr, 'iterations': it, 'batch_size': bs}
 
-    # 训练对率回归模型
+    # 输出最佳参数
-    weights = train_logreg(X_train, y_train, 
+    print(f"Best Parameters: {best_params}")
-                           learning_rate=chosen_learning_rate, 
+    print(f"Best Validation Accuracy: {best_accuracy}")
-                           iterations=chosen_iterations, 
-                           batch_size=chosen_batch_size, 
-                           initial_weights=initial_weights)
 
-    # 测试单个数据
+    # 使用最佳参数训练最终模型
-    y_pred = predict_logreg(X_test[11], weights)
+    final_weights = train_logreg(X_train, y_train, **best_params)
-    print(f"预测结果: {y_pred}, 实际结果: {y_test[11]}")
+    
+    # 在测试集上评估最终模型
+    X_test, y_test = load_data("./data/horse-colic.test")
+    test_accuracy = evaluate_model(X_test, y_test, final_weights)
+    print(f"Test Set Accuracy: {test_accuracy}")