arrow
返回

Encoding High-Cardinality String Categorical Variables

delete2022-03-01
delete59
delete
OA
AI
P
Patricio Cerda *
G
Gaël Varoquaux
DOI:10.1109/TKDE.2020.2992529delete
delete原文链接
delete分享
delete收藏
查看原文
摘要

摘要

En 中文
Statistical models usually require vector representations of categorical variables, using for instance one-hot encoding. This strategy breaks down when the number of categories grows, as it creates high-dimensional feature vectors. Additionally, for string entries, one-hot encoding does not capture morphological information in their representation. Here, we seek low-dimensional encoding of high-cardinality string categorical variables. Ideally, these should be: scalable to many categories; interpretable to end users; and facilitate statistical analysis. We introduce two encoding approaches for string categories: a Gamma-Poisson matrix factorizationon substring counts, and a min-hash encoder, for fast approximation of string similarities. We show that min-hash turns set inclusions into inequality relations that are easier to learn. Both approaches are scalable and streamable. Experiments on real and simulated data show that these methods improve supervised learning with high-cardinality categorical variables. We recommend the following: if scalability is central, the min-hash encoder is the best option as it does not require any data fit; if interpretability is important, the Gamma-Poisson factorization is the best alternative, as it can be interpreted as one-hot encoding on inferred categories with informative feature names. Both models enable autoML on string entries as they remove the need for feature engineering or data cleaning.
Keyword:
Encoding
Statistical analysis
Cleaning
Semantics
Natural language processing
Data models
Machine learning
Statistical learning
string categorical variables
autoML
interpretable machine learning
large-scale data
min-hash
Gamma-Poisson factorization
AI总结

AI总结

对已上传原文的论文进行重点信息的提取,主要内容包括:简要概述、研究摘要、背景介绍、关键亮点、图文解析、展望与总结。

期刊

IEEE Transactions on Knowledge and Data Engineering 封面图
IEEE Transactions on Knowledge and Data Engineering
IF:
10.4
论文数:
7.0K
被引数:
4.1W

机构

I
inria
学者数:
56
论文数: 29
被引数: 2.1K
引用论文

引用论文

The 1965 Eruption of Taal Volcano
err1966-02-25
err0
PREAI
errJames G. Moore; Kazuaki Nakamura; Arturo Alcaraz
err分享
err收藏
err
IF0
err
err0
PREAI
err
err分享
err收藏
EvidenceNOW: Balancing Primary Care Implementation and Implementation Research
err2018-04-09
err0
errOAAI
errDavid Meyers; Therese Miller; Janice Genevro; Chunliu Zhan; Jan De La Mare; Alaina Fournier; Harriet Bennett; Robert J. McNellis
err分享
err收藏
Corticospinal excitation of presumed cervical propriospinal neurones and its reversal to inhibition in humans
err2001-06-01
err0
errOAAI
errGuillaume Nicolas; Véronique Marchand‐Pauvert; David Burke; Emmanuel Pierrot‐Deseilligny
err分享
err收藏
Mössbauer spectroscopical investigation of the exchange biased Fe/MnF2 interface
err2006-10-27
err0
PREAI
errB. Sahoo; W. A. A. Macedo; W. Keune; V. Kuncser; J. Eisenmenger; J. Nogués; I. K. Schuller; I. Felner; Kai Liu; R. Röhlsberger
err分享
err收藏
err分享
err收藏
学者 查看更多内容