决策树ID3预测实例代码

决策树ID3是一种常用的分类算法，它通过计算信息增益来选择最优的特征进行分类。下面是一个简单的决策树ID3预测的实例代码：

import numpy as np

def calc_entropy(data):
    # 计算数据集的熵
    labels = data[:, -1]
    label_counts = np.unique(labels, return_counts=True)
    probs = label_counts[1] / len(labels)
    entropy = -np.sum(probs * np.log2(probs))
    return entropy

def split_data(data, feature_index, feature_value):
    # 根据特征和特征值划分数据集
    mask = data[:, feature_index] == feature_value
    return data[mask]

def choose_best_feature(data):
    # 选择最优的特征
    num_features = data.shape[1] - 1
    base_entropy = calc_entropy(data)
    best_info_gain = 0.0
    best_feature_index = -1

    for i in range(num_features):
        feature_values = np.unique(data[:, i])
        new_entropy = 0.0
        for value in feature_values:
            sub_data = split_data(data, i, value)
            prob = len(sub_data) / len(data)
            new_entropy += prob * calc_entropy(sub_data)
        info_gain = base_entropy - new_entropy
        if info_gain > best_info_gain:
            best_info_gain = info_gain
            best_feature_index = i

    return best_feature_index

def majority_vote(labels):
    # 多数表决，返回出现次数最多的类别
    label_counts = np.unique(labels, return_counts=True)
    max_count_index = np.argmax(label_counts[1])
    return label_counts[0][max_count_index]

def create_decision_tree(data, features):
    # 创建决策树
    labels = data[:, -1]
    if len(np.unique(labels)) == 1:
        return labels[0]
    if len(data[0]) == 1:
        return majority_vote(labels)

    best_feature_index = choose_best_feature(data)
    best_feature = features[best_feature_index]
    decision_tree = {best_feature: {}}
    feature_values = np.unique(data[:, best_feature_index])
    for value in feature_values:
        sub_features = np.delete(features, best_feature_index)
        sub_data = split_data(data, best_feature_index, value)
        decision_tree[best_feature][value] = create_decision_tree(sub_data, sub_features)

    return decision_tree

def predict(decision_tree, features, test_data):
    # 使用决策树进行预测
    root_feature = list(decision_tree.keys())[0]
    root_value = test_data[features.index(root_feature)]
    sub_tree = decision_tree[root_feature][root_value]
    if isinstance(sub_tree, dict):
        return predict(sub_tree, features, test_data)
    else:
        return sub_tree

# 示例数据集
data = np.array([
    [1, 1, 'yes'],
    [1, 1, 'yes'],
    [1, 0, 'no'],
    [0, 1, 'no'],
    [0, 1, 'no']
])

# 特征列表
features = ['feature1', 'feature2']

# 创建决策树
decision_tree = create_decision_tree(data, features)

# 预测新样本
test_data = [1, 0]
prediction = predict(decision_tree, features, test_data)
print("预测结果:", prediction)

这段代码实现了决策树ID3算法的预测过程。首先，通过计算信息增益选择最优的特征进行分类，然后根据特征和特征值划分数据集，递归地创建决策树。最后，使用决策树进行预测。