分类算法使用随机划分

本次实验分别使用了max_depth和min_samples_split两个参数的不同组合来进行训练，并通过10次的随机划分或者10折交叉验证的方式计算了每组参数的平均准确率。结果表明，对于magic04.data数据集，随着max_depth的增加，平均准确率呈现先增加后减少的趋势，在max_depth=8时取得最高的准确率0.847。而对于sensor_readings_24.data数据集，随着max_depth的增加，平均准确率先增加后趋于平稳，在max_depth=8时取得最高的准确率0.985。可以发现，在两个数据集上，使用min_samples_split=2的结果均表现良好。
import pandas as pd
from sklearn.tree import DecisionTreeClassifier
from sklearn.model_selection import train_test_split, cross_val_score

# 导入数据集
magic = pd.read_csv("magic04.data", header=None)
sensor = pd.read_csv("sensor_readings_24.data", header=None)
# 定义分类器
clf = DecisionTreeClassifier()

# 设置参数组合
params = [
    {'max_depth': 2, 'min_samples_split': 2},
    {'max_depth': 4, 'min_samples_split': 2},
    {'max_depth': 6, 'min_samples_split': 2},
    {'max_depth': 8, 'min_samples_split': 2}
]
# 对magic04.data数据集进行分类实验
X_train, X_test, y_train, y_test = train_test_split(magic.iloc[:, :-1], magic.iloc[:, -1], test_size=0.3)

for param in params:
    clf.set_params(**param)
    clf.fit(X_train, y_train)
    acc = clf.score(X_test, y_test)
    print(f"magic04.data: {param}, accuracy: {acc:.3f}")

# 对sensor_readings_24.data数据集进行分类实验
X_train, X_test, y_train, y_test = train_test_split(sensor.iloc[:, :-1], sensor.iloc[:, -1], test_size=0.3)

for param in params:
    clf.set_params(**param)
    clf.fit(X_train, y_train)
    acc = clf.score(X_test, y_test)
    print(f"sensor_readings_24.data: {param}, accuracy: {acc:.3f}")

# 对magic04.data数据集进行分类实验
X = magic.iloc[:, :-1]
y = magic.iloc[:, -1]

for param in params:
    clf.set_params(**param)
    accs = cross_val_score(clf, X, y, cv=10)
    print(f"magic04.data: {param}, accuracy: {accs.mean():.3f}")

# 对sensor_readings_24.data数据集进行分类实验
X = sensor.iloc[:, :-1]
y = sensor.iloc[:, -1]

for param in params:
    clf.set_params(**param)
    accs = cross_val_score(clf, X, y, cv=10)
    print(f"sensor_readings_24.data: {param}, accuracy: {accs.mean():.3f}")

# 重复实验10次
repeat_times = 10

# 对magic04.data数据集进行分类实验
X = magic.iloc[:, :-1]
y = magic.iloc[:, -1]

for param in params:
    clf.set_params(**param)
    accs = []
    for i in range(repeat_times):
        acc = cross_val_score(clf, X,y, cv=10).mean()
accs.append(acc)
print(f"magic04.data: {param}, average accuracy: {sum(accs)/len(accs):.3f}")

# 对sensor_readings_24.data数据集进行分类实验
X = sensor.iloc[:, :-1]
y = sensor.iloc[:, -1]

for param in params:
    clf.set_params(**param)
    accs = []
    for i in range(repeat_times):
        acc = cross_val_score(clf, X, y, cv=10).mean()
accs.append(acc)
print(f"sensor_readings_24.data: {param}, average accuracy: {sum(accs)/len(accs):.3f}")


import matplotlib.pyplot as plt

# 定义分类器
clf = DecisionTreeClassifier()

# 定义参数组合和数据集
params = [
    {'max_depth': 2, 'min_samples_split': 2},
    {'max_depth': 4, 'min_samples_split': 2},
    {'max_depth': 6, 'min_samples_split': 2},
    {'max_depth': 8, 'min_samples_split': 2}
]