Smart Information Spreading for Opinion Maximization in Social Networks.

This repository is based on the paper - Smart Information Spreading for Opinion Maximization in Social Networks.

ArXiv version: https://arxiv.org/pdf/1901.00209

TL;DR:

This paper introduces a novel approach to opinion maximization in social networks by leveraging smart information spreading instead of simply identifying influential nodes. Using a dynamic Bayesian network (DBN), the problem is framed as a sequential decision process, where one source injects information strategically and others randomly. The study develops centralized and decentralized versions of Q-learning algorithms to approximate solutions and, through simulations, shows that smart spreading significantly outperforms random spreading, even when placed at a disadvantageous position in the network.

Notebook: Qopin.ipynb

Installation

Clone the repository:

git clone https://github.com/nayakanuj/Smart_Info_Spreading.git

Navigate to the project directory:

cd Smart_Info_Spreading

Install the required dependencies:

pip install -r requirements.txt

Usage

To run the main script, use the following command:

python qopin_wrapper.py

Results

Comparison between smart and random information spreading in random geometric graph. Smart information spreading is based on a decentralized version of Q-learning.

Comparison between smart and random information spreading in Facebook Ego network [Ref]. Decentralized version of Q-learning algorithm for smart information spreading performs well particularly in clustered graphs.

If found useful, please cite as:

@inproceedings{nayak2019smart,
  title={Smart information spreading for opinion maximization in social networks},
  author={Nayak, Anuj and Hosseinalipour, Seyyedali and Dai, Huaiyu},
  booktitle={IEEE INFOCOM 2019-IEEE Conference on Computer Communications},
  pages={2251--2259},
  year={2019},
  organization={IEEE}
}