Cost-effective Missing Value Imputation for Data-effective Machine Learning

Chengliang Chai; Kaisen Jin; Nan Tang; Ju Fan; Dongjing Miao; Jiayi Wang; Yuyu Luo; Guoliang Li; Ye Yuan; Guoren Wang

doi:10.1145/3716376

Cost-effective Missing Value Imputation for Data-effective Machine Learning

Chengliang Chai
, Kaisen Jin
, Nan Tang
, Ju Fan
, Dongjing Miao
, Jiayi Wang
, Yuyu Luo
, Guoliang Li
, Ye Yuan
, Guoren Wang

计算机学院

Beijing Institute of Technology
The Hong Kong University of Science and Technology (Guangzhou)
Renmin University of China
Faculty of Computing, Harbin Institute of Technology
Tsinghua University

科研成果: 期刊稿件 › 文章 › 同行评审

3 引用（Scopus）

摘要

Given a dataset with incomplete data (e.g., missing values), training a machine learning model over the incomplete data requires two steps. First, it requires a data-effective step that cleans the data in order to improve the data quality (and the model quality on the cleaned data). Second, it requires a data-efficient step that selects a core subset of the data (called coreset) such that the trained models on the entire data and the coreset have similar model quality, in order to save the computational cost of training. The first-data-effective-then-data-efficient methods are too costly, because they are expensive to clean the whole data; while the first-data-efficient-then-data-effective methods have low model quality, because they cannot select high-quality coreset for incomplete data. In this article, we investigate the problem of coreset selection over incomplete data for data-effective and data-efficient machine learning. The essential challenge is how to model the incomplete data for selecting high-quality coreset. To this end, we propose the GoodCore framework towards selecting a good coreset over incomplete data with low cost. To model the unknown complete data, we utilize the combinations of possible repairs as possible worlds of the incomplete data. Based on possible worlds, GoodCore selects an expected optimal coreset through gradient approximation without training ML models. We formally define the expected optimal coreset selection problem, prove its NP-hardness, and propose a greedy algorithm with an approximation ratio. To make GoodCore more efficient, we propose optimization methods that incorporate human-in-the-loop imputation or automatic imputation method into our framework. Moreover, a group-based strategy is utilized to further accelerate the coreset selection with incomplete data given large datasets. Experimental results show the effectiveness and efficiency of our framework with low cost.

源语言	英语
期刊论文编号	10
期刊	ACM Transactions on Database Systems
卷	50
期	3
DOI	https://doi.org/10.1145/3716376
出版状态	已出版 - 14 5月 2025

访问文件

10.1145/3716376

其它文件与链接

链接到 Scopus 的出版物

引用此

@article{a9755e3c16aa415398d1d278a916519b,

title = "Cost-effective Missing Value Imputation for Data-effective Machine Learning",

abstract = "Given a dataset with incomplete data (e.g., missing values), training a machine learning model over the incomplete data requires two steps. First, it requires a data-effective step that cleans the data in order to improve the data quality (and the model quality on the cleaned data). Second, it requires a data-efficient step that selects a core subset of the data (called coreset) such that the trained models on the entire data and the coreset have similar model quality, in order to save the computational cost of training. The first-data-effective-then-data-efficient methods are too costly, because they are expensive to clean the whole data; while the first-data-efficient-then-data-effective methods have low model quality, because they cannot select high-quality coreset for incomplete data. In this article, we investigate the problem of coreset selection over incomplete data for data-effective and data-efficient machine learning. The essential challenge is how to model the incomplete data for selecting high-quality coreset. To this end, we propose the GoodCore framework towards selecting a good coreset over incomplete data with low cost. To model the unknown complete data, we utilize the combinations of possible repairs as possible worlds of the incomplete data. Based on possible worlds, GoodCore selects an expected optimal coreset through gradient approximation without training ML models. We formally define the expected optimal coreset selection problem, prove its NP-hardness, and propose a greedy algorithm with an approximation ratio. To make GoodCore more efficient, we propose optimization methods that incorporate human-in-the-loop imputation or automatic imputation method into our framework. Moreover, a group-based strategy is utilized to further accelerate the coreset selection with incomplete data given large datasets. Experimental results show the effectiveness and efficiency of our framework with low cost.",

keywords = "Data-centric AI, coreset selection, data cleaning, machine learning",

author = "Chengliang Chai and Kaisen Jin and Nan Tang and Ju Fan and Dongjing Miao and Jiayi Wang and Yuyu Luo and Guoliang Li and Ye Yuan and Guoren Wang",

note = "Publisher Copyright: {\textcopyright} 2025 Copyright held by the owner/author(s).",

year = "2025",

month = may,

day = "14",

doi = "10.1145/3716376",

language = "English",

volume = "50",

journal = "ACM Transactions on Database Systems",

issn = "0362-5915",

publisher = "Association for Computing Machinery (ACM)",

number = "3",

}

TY - JOUR

T1 - Cost-effective Missing Value Imputation for Data-effective Machine Learning

AU - Chai, Chengliang

AU - Jin, Kaisen

AU - Tang, Nan

AU - Fan, Ju

AU - Miao, Dongjing

AU - Wang, Jiayi

AU - Luo, Yuyu

AU - Li, Guoliang

AU - Yuan, Ye

AU - Wang, Guoren

PY - 2025/5/14

Y1 - 2025/5/14

N2 - Given a dataset with incomplete data (e.g., missing values), training a machine learning model over the incomplete data requires two steps. First, it requires a data-effective step that cleans the data in order to improve the data quality (and the model quality on the cleaned data). Second, it requires a data-efficient step that selects a core subset of the data (called coreset) such that the trained models on the entire data and the coreset have similar model quality, in order to save the computational cost of training. The first-data-effective-then-data-efficient methods are too costly, because they are expensive to clean the whole data; while the first-data-efficient-then-data-effective methods have low model quality, because they cannot select high-quality coreset for incomplete data. In this article, we investigate the problem of coreset selection over incomplete data for data-effective and data-efficient machine learning. The essential challenge is how to model the incomplete data for selecting high-quality coreset. To this end, we propose the GoodCore framework towards selecting a good coreset over incomplete data with low cost. To model the unknown complete data, we utilize the combinations of possible repairs as possible worlds of the incomplete data. Based on possible worlds, GoodCore selects an expected optimal coreset through gradient approximation without training ML models. We formally define the expected optimal coreset selection problem, prove its NP-hardness, and propose a greedy algorithm with an approximation ratio. To make GoodCore more efficient, we propose optimization methods that incorporate human-in-the-loop imputation or automatic imputation method into our framework. Moreover, a group-based strategy is utilized to further accelerate the coreset selection with incomplete data given large datasets. Experimental results show the effectiveness and efficiency of our framework with low cost.

AB - Given a dataset with incomplete data (e.g., missing values), training a machine learning model over the incomplete data requires two steps. First, it requires a data-effective step that cleans the data in order to improve the data quality (and the model quality on the cleaned data). Second, it requires a data-efficient step that selects a core subset of the data (called coreset) such that the trained models on the entire data and the coreset have similar model quality, in order to save the computational cost of training. The first-data-effective-then-data-efficient methods are too costly, because they are expensive to clean the whole data; while the first-data-efficient-then-data-effective methods have low model quality, because they cannot select high-quality coreset for incomplete data. In this article, we investigate the problem of coreset selection over incomplete data for data-effective and data-efficient machine learning. The essential challenge is how to model the incomplete data for selecting high-quality coreset. To this end, we propose the GoodCore framework towards selecting a good coreset over incomplete data with low cost. To model the unknown complete data, we utilize the combinations of possible repairs as possible worlds of the incomplete data. Based on possible worlds, GoodCore selects an expected optimal coreset through gradient approximation without training ML models. We formally define the expected optimal coreset selection problem, prove its NP-hardness, and propose a greedy algorithm with an approximation ratio. To make GoodCore more efficient, we propose optimization methods that incorporate human-in-the-loop imputation or automatic imputation method into our framework. Moreover, a group-based strategy is utilized to further accelerate the coreset selection with incomplete data given large datasets. Experimental results show the effectiveness and efficiency of our framework with low cost.

KW - Data-centric AI

KW - coreset selection

KW - data cleaning

KW - machine learning

UR - https://www.scopus.com/pages/publications/105009965972

U2 - 10.1145/3716376

DO - 10.1145/3716376

M3 - Article

AN - SCOPUS:105009965972

SN - 0362-5915

VL - 50

JO - ACM Transactions on Database Systems

JF - ACM Transactions on Database Systems

IS - 3

M1 - 10

ER -

Cost-effective Missing Value Imputation for Data-effective Machine Learning

摘要

访问文件

其它文件与链接

学术指纹

引用此