cmpatino/acs-income-2018

该数据集包含来自2018年美国社区调查（ACS）公共使用微数据样本（PUMS）的个体级别记录，专为folktables基准中定义的二元收入预测任务而准备。目标是预测一个人的年总收入是否超过5万美元。数据集包括美国各地的1,611,572个个体，具有10个人口统计、就业和社会经济特征。每个记录都标有一个二元结果，指示个人的总个人收入（PINCP）是否超过每年5万美元。该任务在Ding等人（2021年）的论文中提出，作为使用1994年普查数据的UCI Adult/Census Income数据集的现代替代品，具有更大的样本量、更新的数据（2018年）和更好的地理粒度，适用于机器学习、算法公平性和社会经济分析研究。This dataset contains individual-level records from the 2018 American Community Survey (ACS) Public Use Microdata Sample (PUMS), prepared for the binary income prediction task defined in the folktables benchmark. The goal is to predict whether a persons total annual income exceeds $50,000. The dataset includes 1,611,572 individuals across the United States with 10 demographic, employment, and socioeconomic features. Each record is labeled with a binary outcome indicating whether the individuals total personal income (PINCP) exceeds $50,000 per year. This task was introduced as a modern replacement for the UCI Adult/Census Income dataset (which uses 1994 Census data) in the paper by Ding et al. (2021), offering a larger sample size, current data (2018), and geographic granularity for research in machine learning, algorithmic fairness, and socioeconomic analysis.