长期以来,许多数据集都包含大量缺失数值数据,这一直是一个公认的问题。将机器学习方法应用于数据集时,一个可能至关重要的前提就是解决这一问题。然而,这是一项具有挑战性的任务。本文将近期发展的多层随机优化方法应用于海量医疗记录中的插补问题。该方法基于计算应用数学技术,具有很高的精度。特别是,对于最佳线性无偏预测量(Best Linear Unbiased Predictor,BLUP),这种多层表述是精确的,并且在速度上显著更快、数值上更稳定。这使得 Kriging 方法能够实际应用于海量数据集的数据插补问题。我们在医疗保健研究与质量局(Agency for Healthcare Research and Quality)、医疗保健成本与利用项目(Healthcare Cost and Utilization Project,HCUP)的全国住院患者样本(National Inpatient Sample,NIS)数据记录上测试了该方法。数值结果表明,多层方法显著优于现有方法,并且具有数值稳健性。与 HCUP 近期报告推荐的方法相比,它具有更高的准确性。基准测试显示,误差最多降低 75%。此外,其结果也优于判别式深度学习等近期最先进方法。
英文摘要:
It has long been a recognized problem that many datasets contain significant levels of missing numerical data. A potentially critical predicate for application of machine learning methods to datasets involves addressing this problem. However, this is a challenging task. In this paper, we apply a recently developed multi-level stochastic optimization approach to the problem of imputation in massive medical records. The approach is based on computational applied mathematics techniques and is highly accurate. In particular, for the Best Linear Unbiased Predictor (BLUP) this multi-level formulation is exact, and is significantly faster and more numerically stable. This permits practical application of Kriging methods to data imputation problems for massive datasets. We test this approach on data from the National Inpatient Sample (NIS) data records, Healthcare Cost and Utilization Project (HCUP), Agency for Healthcare Research and Quality. Numerical results show that the multi-level method significantly outperforms current approaches and is numerically robust. It has superior accuracy as compared with methods recommended in the recent report from HCUP. Benchmark tests show up to 75% reductions in error. Furthermore, the results are also superior to recent state of the art methods such as discriminative deep learning.
One Method Does Not Fit All: A Large-Scale Benchmark of Imputation Methods for Single-Cell RNA Sequencing Data
RPTU University Kaiserslautern-Landau, Department of Computer Science(莱茵-瓦尔登堡大学凯撒斯劳滕-兰道分校计算机科学系);German Research Center for Artificial Intelligence (DFKI GmbH)(德国人工智能研究中心(DFKI GmbH));Graduate School of Informatics, Osaka Metropolitan University(大阪 Metropolitan 大学信息研究生院);intelligentX GmbH (intelligentX.com)(intelligentX GmbH(intelligentX.com))
Propensity score estimation using classification and regression trees in the presence of missing covariate data
Leiden University Medical Center(莱顿大学医学中心)
数据挖掘和机器学习技术,如分类和回归树(CART)代表了一个有前途的替代传统的逻辑斯谛回归的倾向得分估计。尽管不完整的数据排除了对所有受试者进行逻辑斯谛回归的拟合,但CART之所以吸引人,部分原因是某些实现允许将不完整的记录并入树拟合中,并为所有受试者提供倾向得分估计。基于理论上的考虑,我们认为,自动处理缺失数据的CART可能是不合适的。使用一系列模拟实验,我们检查了不同方法处理缺失协变量数据的性能;(i)将CART算法直接应用于(部分)不完整的数据,(ii)完整的案例分析,(iii)多重插补。评估了评估确认-结局效应\add{among the exposed}的偏倚、标准误、均方误差和覆盖率。将CART算法直接应用于不完整的数据会导致偏倚,即使在数据完全随机完全随机缺失的情况下也是如此。总体而言,多重插补后加CART的结果最佳。我们的研究表明,在CART中自动处理缺失数据可能会导致严重偏倚,并且作为解释缺失数据的一种手段,其性能并不优于多重插补。
Shapley Value Estimation for Multi-Site Data with Blockwise-Missing Features
Centre for Biomedical Data Science, Duke-NUS Medical School(杜克-新加坡国立大学医学院生物医学数据科学中心);School of Data Science, Chinese University of Hong Kong, Shenzhen(香港中文大学(深圳)数据科学学院);Columbia University(哥伦比亚大学);National University of Singapore(新加坡国立大学);Peking University(北京大学)
Handling Overlapping Asymmetric Datasets -- A Twice Penalized P-Spline Approach
Newcastle University(纽卡斯尔大学);School of Computing(计算机学院);School of Mathematics, Statistics and Physics(数学、统计与物理学院);Translational & Clinical Research Institute(转化与临床研究院)