VAET: A Visual Analytics Approach for E-Transactions Time-Series.

Area of Science:

• Computer Science
• Information Visualization
• Data Mining

Background:

• E-transaction time-series analysis often overlooks significant temporal patterns within large datasets.
• Identifying time-stamped, situation-relevant transactions requires advanced analytical tools.

Purpose of the Study:

• To propose Visual Analysis of E-transaction Time-Series (VAET), an interactive visual analytics system.
• To enable analysts to effectively explore large e-transaction datasets and uncover time-varying transaction insights.

Main Methods:

• Developed VAET, a visual analytics system for interactive exploration of e-transaction time-series.
• Employed a probabilistic decision tree learner to estimate transaction saliency based on analyst-provided training samples.
• Introduced KnotLines for compact visual representation of temporal variations and contextual connections of transactions.

Main Results:

• VAET effectively estimates transaction saliency in large time-series datasets.
• The Time-of-Saliency (TOS) map allows exploration of transactions at various time granularities.
• KnotLines provide an intuitive method for investigating interesting transactions.

Conclusions:

• VAET demonstrates effectiveness in identifying significant time-varying patterns in large e-transaction datasets.
• The system enhances analysts' ability to discover and investigate important transactions through interactive visualization.
• User and case studies confirm VAET's utility with real-world, large-scale e-transaction data.