AppGen: Mobility-aware App Usage Behavior Generation for Mobile Users

TL;DR

This paper tackles privacy-aware data shortages in mobile app usage by introducing AppGen, a mobility-aware generator that synthesizes realistic app usage sequences conditioned on users' spatio-temporal trajectories. It combines an autoregressive diffusion framework with latent representations, a self-attentive historical feature module, and an urban knowledge graph to ground generation in context. Empirical results on two real-world datasets show AppGen outperforming state-of-the-art baselines by over 12% across several fidelity metrics, while preserving spatio-temporal patterns and supporting downstream tasks such as prediction and data augmentation. The work demonstrates that integrating diffusion-based generation with rich contextual encoding yields high-quality synthetic datasets suitable for privacy-preserving analytics and practical applications.

Abstract

Mobile app usage behavior reveals human patterns and is crucial for stakeholders, but data collection is costly and raises privacy issues. Data synthesis can address this by generating artificial datasets that mirror real-world data. In this paper, we propose AppGen, an autoregressive generative model designed to generate app usage behavior based on users' mobility trajectories, improving dataset accessibility and quality. Specifically, AppGen employs a probabilistic diffusion model to simulate the stochastic nature of app usage behavior. By utilizing an autoregressive structure, AppGen effectively captures the intricate sequential relationships between different app usage events. Additionally, AppGen leverages latent encoding to extract semantic features from spatio-temporal points, guiding behavior generation. These key designs ensure the generated behaviors are contextually relevant and faithfully represent users' environments and past interactions. Experiments with two real-world datasets show that AppGen outperforms state-of-the-art baselines by over 12% in critical metrics and accurately reflects real-world spatio-temporal patterns. We also test the generated datasets in applications, demonstrating their suitability for downstream tasks by maintaining algorithm accuracy and order.

Figures (20)

• Figure 1: Finance
• Figure 2: Games
• Figure 3: News
• Figure 5: App category usage records distribution in terms of locations based on Shanghai dataset.
• Figure 6: The graphical framework of the forward and reverse processes of the diffusion models.