更新 function.py

function.py

@@ -1,133 +1,108 @@
 import numpy as np
-from sklearn.impute import SimpleImputer
-from sklearn.feature_selection import SelectKBest, f_classif
 
-# 填写学号和姓名
 id = "202210423213"
 name = "熊浩燃"
 
-def load_data(filepath, k_features=20):
+# 对率回归训练函数，包含加权损失、正则化、学习率调度和提前停止
-    """
+def train_logreg(X, y, learning_rate=0.01, iterations=1000, regularization=0.01, early_stopping_rounds=100, class_weights={0: 1, 1: 1}):
-    从文件中加载数据并进行预处理，包括缺失值填充和特征选择。
+    weights = np.zeros(X.shape[1])
+    min_loss = float('inf')
+    stopping_step = 0
 
-    Args:
+    for i in range(iterations):
-    - filepath (str): 数据文件路径。
+        z = np.dot(X, weights)
-    - k_features (int): 选择的特征数量。
+        predictions = 1 / (1 + np.exp(-z))
 
-    Returns:
+        loss = -np.mean(class_weights[1] * y * np.log(predictions) + 
-    - np.array: 特征数据。
+                        class_weights[0] * (1 - y) * np.log(1 - predictions)) + regularization * np.sum(weights**2)
-    - np.array: 标签数据。
+
-    """
+        if loss < min_loss:
+            min_loss = loss
+            stopping_step = 0
+        else:
+            stopping_step += 1
+        if stopping_step >= early_stopping_rounds:
+            print("提前停止于迭代:", i)
+            break
+
+        gradient = (np.dot(X.T, (predictions - y)) + regularization * weights) / y.size
+        weights -= learning_rate * gradient
+
+        if i % 100 == 0 and i != 0:
+            learning_rate *= 0.9  # 每100轮衰减学习率
+            print(f"迭代 {i}，损失: {loss}，学习率: {learning_rate}")
+
+    return weights
+
+# 对率回归预测函数，带有自定义阈值
+def predict_logreg(inX, weights, threshold=0.6):
+    probability = 1 / (1 + np.exp(-np.dot(inX, weights)))
+    return 1 if probability >= threshold else 0
+
+# 数据加载函数，包含缺失值填充
+def load_data(filepath):
     data = []
     labels = []
     with open(filepath, 'r') as file:
         for line in file:
-            line = line.strip().split()
+            fields = line.strip().split()
-            if line:
+            features = [float(x) if x != '?' else None for x in fields[:-1]]
-                features = []
+            label = float(fields[-1])
-                for feature in line[:-1]:  # 最后一个元素是标签
+            data.append(features)
-                    try:
+            labels.append(label)
-                        features.append(float(feature))
-                    except ValueError:
-                        features.append(np.nan)  # 用 NaN 表示缺失值
-                label = int(line[-1])
-                data.append(features)
-                labels.append(label)
 
-    # 转换为 numpy 数组
+    data = np.array(data, dtype=float)
-    data = np.array(data)
+    labels = np.array(labels, dtype=float)
-    labels = np.array(labels)
 
-    # 使用SimpleImputer填充NaN
+    col_means = np.nanmean(data, axis=0)
-    imputer = SimpleImputer(strategy='mean')
+    indices = np.where(np.isnan(data))
-    data = imputer.fit_transform(data)
+    data[indices] = np.take(col_means, indices[1])
-
-    # 特征标准化，避免除零错误
-    mean = data.mean(axis=0)
-    std_dev = data.std(axis=0)
-    std_dev[std_dev == 0] = 1  # 如果标准差为0，将其设为1，避免除以零
-    data = (data - mean) / std_dev
-
-    # 特征选择
-    selector = SelectKBest(f_classif, k=k_features)
-    data = selector.fit_transform(data, labels)
 
     return data, labels
 
-def sigmoid(z):
+# 暴力搜索函数
-    """
+def brute_force_search(X, y, X_test, y_test, learning_rates, iterations_list, regularizations, class_weights_list, thresholds):
-    Sigmoid激活函数，用于对率回归模型。
+    best_params = {}
+    best_weights = None
+    best_accuracy = 0
 
-    Args:
+    for lr in learning_rates:
-    - z (numpy array): 输入数据。
+        for iters in iterations_list:
+            for reg in regularizations:
+                for cw in class_weights_list:
+                    for threshold in thresholds:
+                        print(f"参数组合: 学习率={lr}, 迭代次数={iters}, 正则化={reg}, 类别权重={cw}, 阈值={threshold}")
+                        weights = train_logreg(X, y, learning_rate=lr, iterations=iters, regularization=reg, class_weights=cw)
+                        y_pred = [predict_logreg(x, weights, threshold=threshold) for x in X_test]
+                        accuracy = np.mean(y_pred == y_test)
+                        print(f"当前组合的测试集准确率: {accuracy}")
 
-    Returns:
+                        if accuracy > best_accuracy:
-    - numpy array: Sigmoid函数输出值。
+                            best_accuracy = accuracy
-    """
+                            best_weights = weights
-    return 1 / (1 + np.exp(-z))
+                            best_params = {'learning_rate': lr, 'iterations': iters, 'regularization': reg, 'class_weights': cw, 'threshold': threshold}
 
-def train_logreg(X, y, initial_lr=0.005, iterations=5000, regularization=0.01, decay=0.0001):
+    print("最佳参数组合:", best_params)
-    """
+    print("最佳准确率:", best_accuracy)
-    对率回归训练函数，包含学习率衰减策略。
+    return best_weights, best_params
 
-    Args:
-    - X (numpy array): 特征数据。
-    - y (numpy array): 标签数据。
-    - initial_lr (float): 初始学习率。
-    - iterations (int): 迭代次数。
-    - regularization (float): 正则化强度。
-    - decay (float): 学习率衰减系数。
-
-    Returns:
-    - numpy array: 训练得到的权重。
-    """
-    m, n = X.shape
-    weights = np.zeros(n)
-    learning_rate = initial_lr
-
-    for i in range(iterations):
-        z = np.dot(X, weights)
-        predictions = sigmoid(z)
-        errors = y - predictions
-        gradient = np.dot(X.T, errors) / m - regularization * weights / m
-        weights += learning_rate * gradient
-
-        # 学习率衰减
-        learning_rate = initial_lr / (1 + decay * i)
-
-    return weights
-
-def predict_logreg(inX, weights, threshold=0.5):
-    """
-    对率回归预测函数，包含可调阈值。
-
-    Args:
-    - inX (numpy array): 输入待分类的样本。
-    - weights (numpy array): 训练得到的权重。
-    - threshold (float): 判定阈值，默认为0.5。
-
-    Returns:
-    - int: 预测的分类标签。
-    """
-    probability = sigmoid(np.dot(inX, weights))
-    return 1 if probability >= threshold else 2
-
-# 测试代码
 if __name__ == "__main__":
-    # 加载数据并选择前20个最重要的特征
+    # 加载训练集
-    X_train, y_train = load_data("./data/horse-colic.data", k_features=20)
+    X_train, y_train = load_data("./data/horse-colic.data")
-    X_test, y_test = load_data("./data/horse-colic.test", k_features=20)
+    X_test, y_test = load_data("./data/horse-colic.test")
 
-    # 训练模型
+    # 调整后的超参数组合
-    weights = train_logreg(X_train, y_train, initial_lr=0.005, iterations=5000, regularization=0.01, decay=0.0001)
+    learning_rates = [0.0001, 0.001, 0.01, 0.05, 0.1]
+    iterations_list = [1000, 2000, 5000, 10000]
+    regularizations = [0.01, 0.1, 1.0, 10.0]
+    class_weights_list = [{0: 1, 1: 2}, {0: 1, 1: 3}, {0: 1, 1: 4}, {0: 1, 1: 5}]
+    thresholds = [0.5, 0.55, 0.6, 0.65, 0.7]
 
-    # 模型评估
+    # 进行暴力搜索找到最佳参数组合
-    correct_predictions = 0
+    best_weights, best_params = brute_force_search(X_train, y_train, X_test, y_test, learning_rates, iterations_list, regularizations, class_weights_list, thresholds)
-    for i in range(len(X_test)):
+    print("最佳权重:", best_weights)
-        y_pred = predict_logreg(X_test[i], weights, threshold=0.5)
-        if y_pred == y_test[i]:
-            correct_predictions += 1
-        print(f"预测结果: {y_pred}, 实际结果: {y_test[i]}")
 
-    accuracy = correct_predictions / len(X_test)
+    # 使用最佳权重和阈值在测试集上评估
-    print(f"测试集准确率: {accuracy:.2f}")
+    y_pred = [predict_logreg(x, best_weights, threshold=best_params['threshold']) for x in X_test]
+    accuracy = np.mean(y_pred == y_test)
+    print(f"最终测试集准确率: {accuracy}")