基于深度强化学习的数据探索性会话自动生成
Auto-generation of Data Exploratory Sessions Based on Deep Reinforcement Learning
汪洋1
作者信息
- 1. 新疆维吾尔自治区烟草公司,新疆 乌鲁木齐 830026
- 折叠
摘要
探索性数据分析(EDA)是一种数据分析方法,旨在通过对数据集进行可视化和摘要统计等方式揭示数据的结构、模式和关系.数据分析人员可通过操作交互式地探索不熟悉的数据集,并为用户提供先导性见解.深度强化学习(DRL)已被证明可以用来解决众多难以解决的人工智能挑战,可尝试将EDA与DRL进行结合,提出了一个名为AEDAS的系统.该系统将EDA建模为一个控制决策问题,从而结合一个新颖的DRL架构来自动生成有说服力的探索性会话,并以EDA笔记本的形式呈现.实验表明,该系统生成的EDA笔记本,可以使用户获得切实有效的先导性见解.
Abstract
Exploratory Data Analysis(EDA)is a data analysis method aimed at revealing the structure,patterns,and relationships in a dataset through visualization and summary statistics.Data analysts can interactively explore unfamiliar datasets through operations and provide users with preliminary insights.Deep Reinforcement Learning(DRL)has been proven to address many difficult Artificial Intelligence challenges.One can attempts to combine the EDA and DRL,proposing a system called AEDAS.The system models EDA as a control decision problem,combining a novel DRL architecture to automatically generate the persuasive exploratory sessions and present them in the form of EDA notebooks.Experiments show that the EDA notebooks generated by the system can provide users with tangible and effective preliminary insights.
关键词
探索性数据分析/深度强化学习框架/控制性问题/探索性会话/EDA笔记本Key words
exploratory data analysis/Deep Reinforcement Learning architecture/control problem/exploratory sessions引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据