📄 中文摘要
随着实时数据收集能力的提升，数据流源变得越来越普遍。处理数据流时，一个主要关注点是概念漂移，即数据分布随时间变化的现象，例如由于环境条件的变化。概念的表示（特征相似行为的静态时期）是适应概念漂移的关键思想。通过测试概念表示与观察窗口的相似性，可以检测到新的或先前出现的重复概念的概念漂移。概念表示是通过元信息特征构建的，这些特征描述了概念行为的各个方面。研究发现，之前提出的概念表示依赖于少量的元信息特征。
数据流中的指纹识别概念：有监督与无监督元信息
📄 English Summary
Fingerprinting Concepts in Data Streams with Supervised and Unsupervised Meta-Information
Streaming data sources are increasingly common due to advancements in real-time data collection capabilities. A significant challenge in managing data streams is concept drift, which refers to changes in the data distribution over time, often caused by variations in environmental conditions. Representing concepts, defined as stationary periods exhibiting similar behaviors, is crucial for adapting to concept drift. By assessing the similarity between a concept representation and a window of observations, it becomes possible to detect concept drift towards a new or previously encountered recurring concept. Concept representations are constructed using meta-information features that describe various aspects of concept behavior. The findings indicate that previously proposed concept representations rely on a limited number of meta-information features.