摘要
迭代决策树(GBDT)属于机器学习算法的一种,由多颗决策树组成,所有树的结论累加起来作为最终答案。该算法表达能力强,可用于大部分回归问题。本文以贵州省遵义市某县负荷及天气数据为基础,结合GBDT算法,对该地区未来8天的日最大负荷进行预测。同时与随机森林和支持向量机两种算法的预测结果对比,结果证明GBDT算法对于短期负荷预测有较好的效果。
Gradient boosted decision trees（GBDT） is a kind of machine learning algorithm,is composed of several decision trees. All the conclusions of the trees sum up as the final answer. The algorithm＇ s expression capability is strong, and can be used in most of the regression problems. Based on the load and weather data of a county in Guizhou Zunyi, combined with GBDT algorithm, this paper forecast the next 8 days of the region＇s largest load. At the same time, it was compared with the random forest algorithm and support vector machine （SVM） algorithm. The results proved that GBDT algorithm has better effect on short-term load forecasting .
出处
《贵州电力技术》
2017年第2期82-84,90,共4页
Guizhou Electric Power Technology
关键词
GBDT
负荷预测
预测
GBDT（ Gradient Boosted Decision Trees）
load forecast
forecast
