Temporal Link Prediction in Social Networks Based on Agent Behavior Synchrony and a Cognitive Mechanism

TL;DR

This work proposes a novel approach to link prediction in temporal networks, extending existing methods with a cognitive mechanism that captures the dynamics of the interactions, by simulating the process of forgetting and strengthening connections depending on the intensity of interactions.

Abstract

Temporality, a crucial characteristic in the formation of social relationships, was used to quantify the long-term time effects of networks for link prediction models, ignoring the heterogeneity of time effects on different time scales. In this work, we propose a novel approach to link prediction in temporal networks, extending existing methods with a cognitive mechanism that captures the dynamics of the interactions. Our approach computes the weight of the edges and their change over time, similar to memory traces in the human brain, by simulating the process of forgetting and strengthening connections depending on the intensity of interactions. We utilized five ground-truth datasets, which were used to predict social ties, missing events, and potential links. We found: (a) the cognitive mechanism enables more accurate capture of the heterogeneity of the temporal effect, leading to an average precision improvement of 9\% compared to baselines with competitive AUC. (b) the local structure and synchronous agent behavior contribute differently to different types of datasets. (c) appropriately increasing the time intervals, which may reduce the negative impact from noise when dividing time windows to calculate the behavioral synchrony of agents, is effective for link prediction tasks.

Figures (7)

• Figure 1: An example of weight between two nodes throughout the time in the CogSNet network with exponential and power functions. Parameters' values set to $\mu = 0.4$, $\theta = 0.1$, and $L = 10$ days michalski2021social.
• Figure 2: Predictive evaluation of methods based on CogSNet when parameters were selected to maximize AUC for (a) edge sampling, (b) event sampling, and (c) future links sampling. The ratio in precision plots determines the complexity of the problem - a higher value indicates more links to predict. The precision plot uses the same colors for methods as the AUC plot.
• Figure 3: Predictive evaluation of methods based on CogSNet when parameters were selected to maximize precision for (a) edge sampling, (b) event sampling, and (c) future links sampling. The ratio in precision plots determines the complexity of the problem -- a higher value indicates more links to predict. The precision plot uses the same colors for methods as the AUC plot.
• Figure 4: The contribution of timescale similarity and local structure similarity in different datasets.
• Figure 5: The impact of CogSNet snapshot time interval on results. The parameters yielding the best AUC were selected for the top row of plots, while for the bottom row, the parameters yielding the best precision were chosen. The selected parameters remained constant, with only the time interval changing.