因果发现数据集ILS

本数据集包含一组用于因果发现和结构学习算法验证的合成数据。在“跨媒体因果推理与决策关键技术研究”项目（项目编号：2021ZD0111700）中 ，该数据集主要被用于测试和评估项目所研发的因果结构学习算法的性能。 本数据集为合成数据，不具备真实的观测时间范围。数据的处理与使用时间为项目执行周期（2021年12月至2024年11月）。数据集内容是基于经典的贝叶斯网络（如Asia、Child等）结构和参数生成的合成样本。本数据集包含一个名为 ILS 的主目录，其中含有24个数据文件。文件名遵循[网络名]_[样本数]_[序号].csv的规则。例如，asia_250_1.csv文件代表基于Asia网络结构生成的、包含250个样本的第1个数据文件。本数据集所有文件均为CSV格式的表格型数据 。由于本次汇交不包含独立的字段说明文件，具体字段（列）的含义需使用者参考原始的、公开的贝叶斯网络结构（如 Asia、Child等）进行理解。This dataset contains a set of synthetic data for the validation of causal discovery and structure learning algorithms. This dataset was primarily used to test and evaluate the performance of the causal structure learning algorithms developed under the project "Research on Key Technologies for Cross-Media Causal Inference and Decision-Making" (Project No.: 2021ZD0111700). This dataset is synthetic data and has no real observational time range. The processing and usage period of the data is the project execution cycle (December 2021 to November 2024). The dataset consists of synthetic samples generated based on the structures and parameters of classic Bayesian networks (e.g., Asia, Child, etc.). This dataset includes a main directory named ILS, which contains 24 data files. The file names follow the rule of [network_name]_[sample_size]_[serial_number].csv. For example, the file asia_250_1.csv represents the first data file generated based on the Asia network structure, containing 250 samples. All files in this dataset are tabular data in CSV format. Since this data submission does not include an independent field description document, the meaning of specific fields (columns) must be understood by users by referring to the original, publicly available Bayesian network structures (e.g., Asia, Child, etc.).

搜集汇总

背景与挑战

背景概述

因果发现数据集ILS是一套合成数据，旨在验证因果发现与结构学习算法的性能。该数据集基于Asia、Child等经典贝叶斯网络生成，包含24个CSV格式文件，总数据量为1.85MB。