A Survey on Behavioral Data Representation Learning

TMLR Paper6938 Authors

Abstract: Behavioral data, reflecting dynamic and complex interactions among entities, are pivotal for advancing multidisciplinary research and practical applications. Effective modeling and representation of behavioral data facilitate enhanced understanding, predictive analytics, and informed decision-making across diverse domains. This paper presents a comprehensive taxonomy of behavioral data representation learning methods, categorized by data modalities: tabular data, event sequences, dynamic graphs, and natural language. Within each category, we further dissect methods based on distinct modeling strategies and capabilities, and provide detailed reviews of their developments. Additionally, we extensively discuss significant downstream applications, datasets, and benchmarks, highlighting their roles in guiding methodological development and evaluating performance. To support further exploration in behavioral data representation learning, we release a continuously maintained repository at [Anonymous GitHub](https://anonymous.4open.science/r/BehavioralDataSurvey) that curates the methods and papers covered in this survey.

Submission Type: Long submission (more than 12 pages of main content)

Assigned Action Editor: ~Devendra_Singh_Dhami1

Submission Number: 6938