A Probabilistic Graph Model for Trust Opinion Estimation in Online Social Networks

Abstract

Trust assessment plays a key role in many online applications, such as online money lending, product reviewing and active friending. Trust models usually employ a group of parameters to represent the trust relation between a trustor-trustee pair. These parameters are originated from the trustors bias and opinion on the trustee. Naturally, these parameters can be regarded as a vector. To address this problem, we propose a framework to accurately convert the single values to the parameters needed by 3VSL. The framework firstly employs a probabilistic graph model (PGM) to derive the trustors opinion and bias to his rating on the trustee.

TrustGNN: Graph Neural Network based Trust Evaluation via Learnable Propagative and Composable Nature

“…They adopt bidirectional interaction relationships in online social networks to deconstruct users' social behaviors and apply principal component analysis to estimate interpersonal trust. Liu et al [12] propose a framework to accurately model single values as parameters. The model first uses a probabilistic graph model to model trustors' opinions and biases to his rating on the trustee.…”

“…E.g., the subjective logic-based methods [7]- [9], which follow the assumptions of cognitive recognition and introduce the uncertainty inference process for the subjective nature of trust. The probability statistics-based methods [10]- [12], rely on statistical distributions to represent and model social trust relationship in a computational way. The machine learning-based methods [3], [13], use some machine learning techniques such as matrix factorization to model the trust evaluation task as a learnable problem.…”