加载中
DART-Dropouts Multiple Additive Regression Trees

一句话简述: (将dropouts思想引入MART中,在每棵树的迭代过程中不再单单去拟合前一棵树的残差,而是从前面已有的树中sample一部分树,ensemble一个新model,然后去拟合这部分的残差,从而使...

2018/08/21 15:30
20
xgboost/gbdt在调参时为什么树的深度很少就能达到很高的精度?

问题: 用xgboost/gbdt在在调参的时候把树的最大深度调成6就有很高的精度了。但是用DecisionTree/RandomForest的时候需要把树的深度调到15或更高。用RandomForest所需要的树的深度和Decisio...

2018/08/19 22:49
14
XGboost数据比赛实战之调参篇(完整流程)

这一篇博客的内容是在上一篇博客Scikit中的特征选择,XGboost进行回归预测,模型优化的实战的基础上进行调参优化的,所以在阅读本篇博客之前,请先移步看一下上一篇文章。 我前面所做的工作基...

2018/08/18 15:36
62
第一次参加Kaggle拿银总结

转自:第一次参加Kaggle拿银总结 作者:ScarletPan 我的比赛代码已经放到github --> Kaggle-Rental-Listing-Inquireies 在这篇博客开始之前,我必须感谢导师给我提供服务器资源,@Fenix Lin学...

2018/08/08 00:13
26
特征工程-知识汇总

1 特征工程是什么? 2 数据预处理   2.1 无量纲化     2.1.1 标准化     2.1.2 区间缩放法     2.1.3 标准化与归一化的区别   2.2 对定量特征二值化   2.3 对定性特征哑编...

2018/08/07 23:40
17
资源 | Feature Tools:可自动构造机器学习特征的Python库

作者:William Koehrsen 来源:Towards data science、机器之心 机器学习越来越多地从人工设计模型转向使用 H20、TPOT 和 auto-sklearn 等工具自动优化的工具。这些库以及随机搜索(参见《R...

2018/08/07 23:37
100
Avoid Overfitting By Early Stopping With XGBoost

Overfitting is a problem with sophisticated non-linear learning algorithms like gradient boosting. In this post you will discover how you can use early stopping to limit overfit...

【集成学习】lightgbm调参案例

lightgbm使用leaf_wise tree生长策略,leaf_wise_tree的优点是收敛速度快,缺点是容易过拟合。 # lightgbm关键参数 # lightgbm调参方法cv 代码github地址 1 # -*- coding: utf-8 -*- 2 """ ...

2018/08/05 18:00
161
Home Credit Default Risk-Start Here

https://www.kaggle.com/willkoehrsen/start-here-a-gentle-introduction

2018/08/05 17:58
89
Kaggle:Home Credit Default Risk 特征工程构建及可视化(2)

博主在之前的博客 Kaggle:Home Credit Default Risk 数据探索及可视化(1) 中介绍了 Home Credit Default Risk 竞赛中一个优秀 kernel 关于数据的探索及可视化的工作,本篇博客将围绕如何构...

2018/08/05 17:36
130
Automated Feature Engineering Basics

https://www.kaggle.com/willkoehrsen/automated-feature-engineering-basics

2018/08/05 17:15
5
One Hot Encoding vs LabelEncoder?

There are some cases where LabelEncoder or DictVectorizor are useful, but these are quite limited in my opinion due to ordinality. LabelEncoder can turn [dog,cat,dog,mouse,cat] ...

2018/08/05 17:15
23
Kaggle:Home Credit Default Risk 数据探索及可视化(1)

最近博主在做个 kaggle 竞赛,有个 Kernel 的数据探索分析非常值得借鉴,博主也学习了一波操作,搬运过来借鉴,原链接如下: https://www.kaggle.com/willkoehrsen/start-here-a-gentle-intr...

2018/08/05 16:42
150
Machine Learning Kaggle Competition Part Two

Machine Learning Kaggle Competition Part Two: Improving Feature engineering, feature selection, and model evaluation Like most problems in life, there are several potential appr...

2018/08/05 16:41
17
Introduction to Manual Feature Engineering

https://www.kaggle.com/willkoehrsen/introduction-to-manual-feature-engineering

2018/08/05 16:34
2
How LightGBM handles missing values?

https://www.kaggle.com/c/home-credit-default-risk/discussion/57918 How LightGBM handles missing values? posted in Home Credit Default Risk 2 months ago 8 Hi everyone Can anyone ...

2018/08/05 12:44
113
LightGBM and XGBoost Explained

The gradient boosting decision tree (GBDT) is one of the best performing classes of algorithms in machine learning competitions. One implementation of the gradient boosting deci...

2018/08/05 12:39
66

没有更多内容

加载失败,请刷新页面

没有更多内容

返回顶部
顶部