技术问题梳理

1. 过拟合vs.欠拟合  // 方差vs.偏差

理解机器学习中的偏差与方差：https://blog.csdn.net/simple_the_best/article/details/71167786

理解过拟合：https://blog.csdn.net/SIGAI_CSDN/article/details/80730301

 

2.R数据预处理、缺失值填补

http://www.cnblogs.com/DianaLi/p/9141753.html

 

3.类别不平衡之过采样SMOTE算法

https://www.cnblogs.com/Determined22/p/5772538.html

http://m.elecfans.com/article/620100.html

 R是用DMwR里的SMOTE函数实现，caret包里的createDataPartition是分层

code：https://blog.csdn.net/jiabiao1602/article/details/42392377

perc.over含义：https://blog.csdn.net/hongjinlongno1/article/details/70226683

 

4.信用卡评分

https://mp.weixin.qq.com/s?__biz=MzI0Mzk2NDEwOQ==&mid=2247484059&idx=1&sn=0f2f10637d9c9d9bcd37b567e559b607&chksm=e9644339de13ca2fbc7fb29ba7d387e5d5099afa3d887e5a01ff637a71520866fad1c7d1b4f1&mpshare=1&scene=1&srcid=0319lLTICG9WEWyBmcLVeNz2#rd

 

https://mp.weixin.qq.com/s?__biz=MzA3MTM3NTA5Ng==&mid=2651058871&idx=2&sn=517429f878da1d4b56be020b1e2fb740&chksm=84d9d320b3ae5a36a5f2508419f04c3d0f6237b069202098f0615d93d3ba9021183fd23d7887&mpshare=1&scene=1&srcid=0319l733EJ1OukjYZg41YEfH#rd

 

https://www.jianshu.com/p/159f381c661d

 

https://blog.csdn.net/zpxcod007/article/details/80118580

 

https://blog.csdn.net/ooxxshaso/article/details/79843832

 

https://www.cnblogs.com/nxld/p/6364966.html

 

 

面试考点：

https://mp.weixin.qq.com/s?__biz=MzA3OTAxMDQzNQ==&mid=2650608859&idx=1&sn=43867940bee0f5237414fdf7868dcff5&chksm=87b3bb37b0c43221f39c901f5243185e225af32ce098b6fb6dd7ed9a0cc155b373244d221c6e&mpshare=1&scene=1&srcid=0312ymgMvXumYs9uad7ZppnF#rd

 

特征选择：

https://www.cnblogs.com/wkslearner/p/8933685.html

 

特征选择之信息增益：

https://www.cnblogs.com/mfrbuaa/p/3931706.html