https://plumephp.com/tags/lightgbm/Recent content in LightGBM on PlumePHPHugozh-CNMon, 28 Sep 2026 10:00:00 +0800https://plumephp.com/ai-time-series-forecasting/Sun, 27 Sep 2026 12:30:00 +0800https://plumephp.com/ai-time-series-forecasting/时序预测实战：ARIMA/ETS、LightGBM 特征工程、LSTM/TCN/Transformer、多步预测与生产部署。https://plumephp.com/ml-ensemble-learning/Mon, 28 Sep 2026 10:00:00 +0800https://plumephp.com/ml-ensemble-learning/<h2 id="引言">引言</h2> <p>单个模型（决策树、逻辑回归）常有力不从心的时候：树容易过拟合、线性模型学不动非线性。**集成学习（Ensemble）**把多个「弱模型」组合成一个「强模型」，是几乎所有机器学习竞赛与工业落地的默认武器。本文从偏差-方差视角讲清「为什么集成有效」，再逐步落地 Bagging → 随机森林 → AdaBoost → GBDT → XGBoost/LightGBM → Stacking，全程附 sklearn 可运行代码。</p>