wepe / MachineLearningEveryday

record and share my reading everyday

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

##2016 阅读记录

  • 2016-03-05

Xgboost中两个我之前没有用过的特性,一个是用户自定义代价函数,另一个是pred_leaf,预测的时候设置pred_leaf为True,将对每个样本返回其在每棵树上的leaf index(一共有num_round棵树),可以当成新的特征来用。

  • 2016-03-07

阅读了《A Programmer's Guide to Data Mining》这本电子书,内容过于简单,两三个小时读完,干货不多。

  • 2016-03-08

Converting categorical data into numbers with Pandas and Scikit-learn 这篇文章讲了类别特征的处理,文中提到一点需要引起思考:If you have missing values in a binary feature, there’s an alternative representation:-1 for negatives,0 for missing values,1 for positives。It worked better in case of the Analytics Edge competition: an SVM trained on one-hot encoded data with d indicators scored 0.768 in terms of AUC, while the alternative representation yielded 0.778.

  • 2016-03-11

Imbalanced data – Finding Waldo 这篇文章讲了不平衡数据的处理,都是常见的方法(简单采样,合成采样),但是文章最后讲了一个很有趣的处理方式:如果不平衡数据中某个类别的数据非常少,那么也可以把分类问题当成异常值检测的问题( anomaly detection),只需要检测出异常值就行了。

  • 2016-04-01

看了large-scale svm相关的内容,发现一个不错的工具EnsembleSVM,在准确率不下降的同时减小计算复杂度,对应论文EnsembleSVM: A Library for Ensemble Learning Using Support Vector Machines,另外一篇cite比较多的论文Making large-scale support vector machine learning practical

About

record and share my reading everyday