Efficient Detection of Hot Span in Information Diffusion from Observation

TL;DR

This paper presents an efficient algorithm for detecting change points in information diffusion behavior from limited data, significantly improving accuracy and speed over naive methods.

Contribution

The authors introduce a novel likelihood derivative-based search algorithm that efficiently identifies change patterns without iterative parameter optimization.

Findings

The proposed algorithm accurately detects change points in real-world networks.

It outperforms naive exhaustive search methods in both accuracy and computational efficiency.

Experimental results confirm the method's effectiveness on multiple network structures.

Abstract

We addressed the problem of detecting the change in behavior of information diffusion from a small amount of observation data, where the behavior changes were assumed to be effectively reflected in changes in the diffusion parameter value. The problem is to detect where in time and how long this change persisted and how big this change is. We solved this problem by searching the change pattern that maximizes the likelihood of generating the observed diffusion sequences. The naive learning algorithm has to iteratively update the patten boundaries, each requiring optimization of diffusion parameters by the EM algorithm, and is very inefficient. We devised a very efficient search algorithm using the derivative of likelihood which avoids parameter value optimization during the search. The results tested using three real world network structures confirmed that the algorithm can efficiently identify the correct change pattern. We further compared our algorithm with the naive method that finds the best combination of change boundaries by an exhaustive search through a set of randomly selected boundary candidates, and showed that the proposed algorithm far outperforms the native method both in terms of accuracy and computation time.