CRITEO-UPLIFTv2

CRITEO-UPLIFTv2数据集是由Criteo AI Lab创建的大规模基准数据集，用于个体治疗效果预测和提升建模。该数据集包含13979592个样本，是从多个随机对照试验中收集的，显著扩大了现有数据集的规模。数据集的特点包括治疗不平衡、二元和连续的匿名特征以及低结果率。数据集的创建过程涉及从在线控制实验（A/B测试）中收集数据，以更好地研究广告对点击和销售的个体影响。该数据集的应用领域包括医疗保健、在线广告和经济社会学，旨在通过大规模数据推动因果推断领域的研究，特别是在个体治疗效果预测和提升建模方面。The CRITEO-UPLIFTv2 dataset is a large-scale benchmark dataset created by Criteo AI Lab for individual treatment effect prediction and uplift modeling. It contains 13,979,592 samples collected from multiple randomized controlled trials, which significantly expands the scale of existing datasets. The dataset features imbalanced treatment assignment, binary and continuous anonymized features, as well as a low outcome rate. The dataset was constructed by collecting data from online controlled experiments (A/B tests) to better investigate the individual impacts of advertising on clicks and sales. Its application domains include healthcare, online advertising and economic sociology, aiming to advance causal inference research through large-scale data, particularly in the areas of individual treatment effect prediction and uplift modeling.

搜集汇总

背景与挑战

背景概述

CRITEO-UPLIFTv2数据集是由Criteo AI Lab创建的大规模基准数据集，用于个体治疗效果预测和提升建模，包含约1398万个样本，从多个随机对照试验收集，具有治疗不平衡、匿名特征和低结果率等特点。该数据集基于在线A/B测试数据，研究广告对点击和销售的个体影响，应用领域包括医疗保健、在线广告和经济社会学，旨在推动因果推断领域的研究。