机器学习实战：从吴恩达课程到房价预测项目（Python + Scikit-learn）

机器学习实战从吴恩达课程到房价预测项目Python Scikit-learn1. 项目背景与目标房价预测是机器学习入门的经典案例也是吴恩达机器学习课程中重点讲解的监督学习应用场景。不同于课程中使用的Octave实现本教程将完全基于Python生态使用Scikit-learn等现代工具库带你完成从数据探索到模型部署的全流程。为什么选择房价预测作为实战项目数据维度丰富面积、房龄、地段等适合演示特征工程技巧问题定义清晰回归任务便于验证模型效果业务价值直观预测结果可直接用于实际决策2. 环境准备与数据加载2.1 工具链配置推荐使用Anaconda创建独立环境conda create -n house_price python3.8 conda activate house_price pip install pandas scikit-learn matplotlib seaborn2.2 数据集介绍使用Kaggle的House Prices数据集包含1460条房屋销售记录81个特征字段。关键字段包括import pandas as pd data pd.read_csv(train.csv) print(data.columns.tolist()[:10]) # 查看前10个特征输出示例[Id, MSSubClass, MSZoning, LotFrontage, LotArea, Street, Alley, LotShape, LandContour, Utilities]2.3 数据质量检查# 缺失值统计 missing data.isnull().sum().sort_values(ascendingFalse) missing missing[missing 0] print(missing)典型问题处理方案连续变量缺失中位数填充分类变量缺失单独标记为Missing高缺失率特征80%直接剔除3. 特征工程实战3.1 数值特征处理from sklearn.impute import SimpleImputer from sklearn.preprocessing import StandardScaler # 处理缺失值 num_features data.select_dtypes(include[int64, float64]).columns num_imputer SimpleImputer(strategymedian) data[num_features] num_imputer.fit_transform(data[num_features]) # 标准化处理 scaler StandardScaler() data[num_features] scaler.fit_transform(data[num_features])3.2 类别特征编码from sklearn.preprocessing import OneHotEncoder cat_features data.select_dtypes(include[object]).columns encoder OneHotEncoder(handle_unknownignore, sparseFalse) encoded_cats encoder.fit_transform(data[cat_features]) # 合并处理后的特征 processed_data pd.concat([ data[num_features], pd.DataFrame(encoded_cats, columnsencoder.get_feature_names_out()) ], axis1)3.3 特征相关性分析import seaborn as sns import matplotlib.pyplot as plt corr_matrix processed_data.corr() plt.figure(figsize(12,10)) sns.heatmap(corr_matrix, cmapcoolwarm) plt.title(Feature Correlation Matrix) plt.show()4. 模型构建与优化4.1 基础线性回归from sklearn.linear_model import LinearRegression from sklearn.model_selection import train_test_split X processed_data.drop(SalePrice, axis1) y processed_data[SalePrice] X_train, X_test, y_train, y_test train_test_split(X, y, test_size0.2) model LinearRegression() model.fit(X_train, y_train)4.2 正则化改进from sklearn.linear_model import Ridge ridge Ridge(alpha1.0) ridge.fit(X_train, y_train) # 对比模型表现 print(Linear Regression R2:, model.score(X_test, y_test)) print(Ridge Regression R2:, ridge.score(X_test, y_test))4.3 交叉验证调参from sklearn.model_selection import GridSearchCV param_grid {alpha: [0.001, 0.01, 0.1, 1, 10, 100]} grid_search GridSearchCV(Ridge(), param_grid, cv5) grid_search.fit(X_train, y_train) print(Best alpha:, grid_search.best_params_)5. 模型评估与可视化5.1 评估指标计算from sklearn.metrics import mean_squared_error, r2_score predictions grid_search.best_estimator_.predict(X_test) rmse np.sqrt(mean_squared_error(y_test, predictions)) r2 r2_score(y_test, predictions) print(fRMSE: {rmse:.2f}) print(fR2 Score: {r2:.2f})5.2 残差分析residuals y_test - predictions plt.figure(figsize(10,6)) sns.scatterplot(xpredictions, yresiduals) plt.axhline(y0, colorr, linestyle--) plt.title(Residual Plot) plt.xlabel(Predicted Values) plt.ylabel(Residuals) plt.show()5.3 特征重要性coef pd.Series(grid_search.best_estimator_.coef_, indexX.columns) important_features coef.abs().sort_values(ascendingFalse)[:10] important_features.plot(kindbarh) plt.title(Top 10 Important Features) plt.show()6. 项目进阶方向6.1 集成方法尝试from sklearn.ensemble import RandomForestRegressor rf RandomForestRegressor(n_estimators100, random_state42) rf.fit(X_train, y_train) print(RF R2:, rf.score(X_test, y_test))6.2 自动化机器学习!pip install autosklearn import autosklearn.regression automl autosklearn.regression.AutoSklearnRegressor(time_left_for_this_task120) automl.fit(X_train, y_train) print(automl.leaderboard())6.3 模型部署示例使用Flask创建预测APIfrom flask import Flask, request, jsonify import pickle app Flask(__name__) model pickle.load(open(model.pkl,rb)) app.route(/predict, methods[POST]) def predict(): data request.get_json() prediction model.predict([data[features]]) return jsonify({price: prediction[0]}) if __name__ __main__: app.run(port5000)7. 常见问题解决方案问题1特征维度爆炸方案使用PCA降维或L1正则化筛选特征from sklearn.decomposition import PCA pca PCA(n_components0.95) X_pca pca.fit_transform(X)问题2非线性关系处理方案添加多项式特征from sklearn.preprocessing import PolynomialFeatures poly PolynomialFeatures(degree2, include_biasFalse) X_poly poly.fit_transform(X[[GrLivArea, OverallQual]])问题3类别不平衡方案目标变量对数变换y_log np.log1p(y)在实际项目中我发现特征工程阶段花费的时间往往超过模型构建本身。特别是在处理房屋数据时如何合理组合特征如将地下室面积与地上面积比值作为新特征能显著提升模型表现。另外使用Pipeline可以大幅提升代码的可维护性from sklearn.pipeline import make_pipeline pipeline make_pipeline( SimpleImputer(strategymedian), StandardScaler(), Ridge(alpha1.0) ) pipeline.fit(X_train, y_train)