Pairwise-based visual labeling system for event-tracking data(基于成对关系的埋点数据可视化标注系统)

Abstract

Event-tracking technology embeds special code within applications to collect data during runtime, which aids developers in performance analysis and evaluation. Multidimensional event-tracking data is often automatically analyzed through evaluation models to alleviate manual labor. However, labeling event-tracking data, a crucial step in model training, remains laborious and inefficient. This paper presents a novel pairwise-based event-tracking data labeling method and an interactive visual labeling system to enhance the labeling process. We demonstrate the effectiveness and utility of the proposed method and system through a case study and a user study.(埋点技术，即利用应用程序中嵌入的特殊代码收集运行过程中的数据，方便开发者分析和评估性能。多维埋点数据通常借助评估模型进行自动化分析，以减轻人工负担。然而，埋点数据标注作为训练模型的重要环节，仍存在难度大、效率低的问题。设计了一种基于成对关系的埋点数据标注方法，实现了一个交互式可视化标注系统，有效降低了埋点数据的标注难度，减轻了标注负担，并显著提升了标注效率。通过基于真实数据的案例分析和用户实验，验证了方法和系统的有效性和实用性。)

Keywords

• event-tracking data(埋点数据)
• interactive(交互式)
• visual analytics(可视分析)
• data labeling(数据标注)
• pairwise(成对关系)