🧠 一、开始
真正动手实现一个完整的AI项目!从数据预处理、特征工程、模型训练,到评估与调优,一步步还原你在动画视频中看到的所有核心知识点。
📦 二、环境准备
建议使用 Python 3.8+,推荐工具:
shell
pip install scikit-learn pandas matplotlib seaborn numpy
📊 三、案例背景:预测大学生能否被录取
我们将使用一个简化的数据集(模拟大学申请系统),包含:
GRE Score | TOEFL Score | GPA | University Rating | Research | Admit |
---|---|---|---|---|---|
330 | 115 | 9.0 | 5 | 1 | 1 |
312 | 103 | 8.1 | 3 | 0 | 0 |
... | ... | ... | ... | ... | ... |
📥 四、加载数据与可视化探索
python
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt
df = pd.read_csv('https://raw.githubusercontent.com/amankharwal/Website-data/master/Admission_Predict.csv')
df.columns = df.columns.str.strip()
# 简化目标变量
df['Admit'] = df['Chance of Admit '] > 0.75
df['Admit'] = df['Admit'].astype(int)
# 预览数据
print(df.head())
# 可视化相关性
sns.heatmap(df.corr(), annot=True, cmap='Blues')
plt.title('Correlation Matrix')
plt.show()
🧹 五、特征工程 & 数据预处理
python
from sklearn.model_selection import train_test_split
from sklearn.preprocessing import StandardScaler
X = df.drop(['Chance of Admit ', 'Admit'], axis=1)
y = df['Admit']
# 标准化
scaler = StandardScaler()
X_scaled = scaler.fit_transform(X)
# 拆分训练集与测试集
X_train, X_test, y_train, y_test = train_test_split(X_scaled, y, test_size=0.2, random_state=42)
🤖 六、构建模型与训练
我们使用逻辑回归(Logistic Regression)作为分类模型,后续也会加入其他模型做对比。
python
from sklearn.linear_model import LogisticRegression
from sklearn.metrics import accuracy_score, classification_report
model = LogisticRegression()
model.fit(X_train, y_train)
# 预测
y_pred = model.predict(X_test)
print(f'Accuracy: {accuracy_score(y_test, y_pred):.4f}')
print(classification_report(y_test, y_pred))
🔍 七、可视化决策边界(仅用于2D简化版)
python
import numpy as np
# 降维为两个特征(仅用于可视化)
X2D = X_scaled[:, :2]
X_train2D, X_test2D, _, _ = train_test_split(X2D, y, test_size=0.2, random_state=42)
model2D = LogisticRegression()
model2D.fit(X_train2D, y_train)
# 生成边界图
x_min, x_max = X2D[:, 0].min() - 1, X2D[:, 0].max() + 1
y_min, y_max = X2D[:, 1].min() - 1, X2D[:, 1].max() + 1
xx, yy = np.meshgrid(np.arange(x_min, x_max, 0.1),
np.arange(y_min, y_max, 0.1))
Z = model2D.predict(np.c_[xx.ravel(), yy.ravel()])
Z = Z.reshape(xx.shape)
plt.contourf(xx, yy, Z, alpha=0.4)
plt.scatter(X2D[:, 0], X2D[:, 1], c=y, s=20, edgecolor='k')
plt.title('Decision Boundary (2 Features)')
plt.xlabel('GRE Score')
plt.ylabel('TOEFL Score')
plt.show()
🧪 八、模型评估与过拟合检测
我们加入其他模型(随机森林)对比训练集与测试集效果,检测过拟合。
python
from sklearn.ensemble import RandomForestClassifier
forest = RandomForestClassifier(n_estimators=100, random_state=42)
forest.fit(X_train, y_train)
train_score = forest.score(X_train, y_train)
test_score = forest.score(X_test, y_test)
print(f'RandomForest 训练集准确率: {train_score:.4f}')
print(f'RandomForest 测试集准确率: {test_score:.4f}')
训练集准确率远高于测试集 => 可能过拟合
✅ 九、正则化概念演示(L1, L2)
python
from sklearn.linear_model import LogisticRegressionCV
lr_cv = LogisticRegressionCV(cv=5, penalty='l2', solver='liblinear')
lr_cv.fit(X_train, y_train)
print("最佳C值:", lr_cv.C_[0])
🧠 十、关键概念总结(边看边实践)
概念 | 示例代码 | 对应含义 |
---|---|---|
特征工程 | StandardScaler |
统一数值尺度 |
模型训练 | LogisticRegression().fit() |
学习数据规律 |
过拟合 | 训练准确率高但测试低 | 模型记忆训练集 |
正则化 | penalty='l2' |
抑制复杂模型 |
模型选择 | Logistic, RandomForest |
尝试多个模型对比 |
🚀 Bonus:自动化训练多个模型对比性能
python
from sklearn.model_selection import cross_val_score
from sklearn.svm import SVC
from sklearn.naive_bayes import GaussianNB
models = {
'Logistic': LogisticRegression(),
'RandomForest': RandomForestClassifier(),
'SVM': SVC(),
'NaiveBayes': GaussianNB()
}
for name, model in models.items():
scores = cross_val_score(model, X_scaled, y, cv=5)
print(f'{name}: 平均准确率 {scores.mean():.4f}')