On fast and scalable recurring link’s prediction in evolving multi-graph streams

Abstract

The link prediction task has found numerous applications in real-world scenarios. However, in most of the cases like interactions, purchases, mobility, etc., links can re-occur again and again across time. As a result, the data being generated is excessively large to handle, associated with the complexity and sparsity of networks. Therefore, we propose a very fast, memory-less, and dynamic sampling-based method for predicting recurring links for a successive future point in time. This method works by biasing the links exponentially based on their time of occurrence, frequency, and stability. To evaluate the efficiency of our method, we carried out rigorous experiments with massive real-world graph streams. Our empirical results show that the proposed method outperforms the state-of-the-art method for recurring links prediction. Additionally, we also empirically analyzed the evolution of links with the perspective of multi-graph topology and their recurrence probability over time.

Forgetting on Evolving Graphs for Accurate and Diverse Stream-Based Recommendation

“…We use IGSI πt without forgetting, and also extend it with both our proposed forgetting technique and the following ones: Sliding window -A traditional abrupt technique. We define windows based on time intervals of size τ , where observations outside of the windows are forgotten; Time-based decay -A technique [Tabassum et al 2020] that sporadically reduces relevance of edges based on time.…”

“…Alternatively, gradual forgetting relies on mechanisms to weight observations based on recency, by considering recent to be more important than older ones. Techniques to gradually decrease importance of observations over time and assign higher weights to more recent ones have been explored in neighborhood-based methods [Ding and Li 2005;Tabassum et al 2020]. In general, recommendations based on recent information provide improved accuracy.…”

Introduction to the special issue on COMPLEX NETWORKS 2018

“…Their study highlights the need to develop models that are able to extract and quantify the variability existing in network populations. Tabassum et al (2020) proposed an algorithm for predicting recurring links in evolving weighted multigraph. While previous works mainly rely on proximity measures or structural heuristics that are computationally expensive for massive graph streams (Marjan et al, 2018), this dynamic sampling-based method predicts recurring links without exploiting the time and space expensive neighborhood for each node/link, or global information of the network.…”