Can We Enhance the Quality of Mobile Crowdsensing Data Without Ground Truth?

2024-05-29

arXiv.org

AI Summary

To address the data quality enhancement challenge in mobile crowdsourcing (MCS) under label-free scenarios, this paper proposes PRBTD—a joint prediction-and-reputation-based truth discovery framework. PRBTD innovatively integrates spatiotemporal Transformer-based prediction of latent ground truths with implicit data correlation modeling derived from predictions, enabling dynamic calibration of user reputation and data truthfulness without any ground-truth annotations—thereby overcoming reliance on prior assumptions or manual labeling inherent in conventional truth discovery methods. Leveraging correlation-driven spatiotemporal modeling and reputation-weighted iterative optimization, PRBTD achieves state-of-the-art performance across multiple real-world MCS datasets: it improves malicious user detection accuracy by 12.6%, attains a 94.3% recall rate for low-quality data, and consistently enhances downstream task accuracy.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Humans and AI: Crowd Sourcing and Human Computation
• Data Mining & Knowledge Management: Mining of Spatial, Temporal or Spatio-Temporal Data

Application Category

• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Mobile crowdsensing (MCS) has emerged as a prominent trend across various domains. However, ensuring the quality of the sensing data submitted by mobile users (MUs) remains a complex and challenging problem. To address this challenge, an advanced method is needed to detect low-quality sensing data and identify malicious MUs that may disrupt the normal operations of an MCS system. Therefore, this article proposes a prediction- and reputation-based truth discovery (PRBTD) framework, which can separate low-quality data from high-quality data in sensing tasks. First, we apply a correlation-focused spatio-temporal Transformer network that learns from the historical sensing data and predicts the ground truth of the data submitted by MUs. However, due to the noise in historical data for training and the bursty values within sensing data, the prediction results can be inaccurate. To address this issue, we use the implications among the sensing data, which are learned from the prediction results but are stable and less affected by inaccurate predictions, to evaluate the quality of the data. Finally, we design a reputation-based truth discovery (TD) module for identifying low-quality data with their implications. Given the sensing data submitted by MUs, PRBTD can eliminate the data with heavy noise and identify malicious MUs with high accuracy. Extensive experimental results demonstrate that the PRBTD method outperforms existing methods in terms of identification accuracy and data quality enhancement.

Problem

• Mobile Crowd Sensing
• Data Accuracy
• Malicious Interference

Innovation

• PRBTD Framework
• Data Quality Enhancement
• Malicious Behavior Detection