Balanced Popularity in Multi-Product Billboard Advertisement

TL;DR

This paper addresses the challenge of selecting billboard advertisement slots for multiple products to maximize overall influence while maintaining balanced popularity, proposing algorithms and heuristics validated by real-world data.

Contribution

It formulates a novel multi-product influence maximization problem with balanced influence constraints and develops linear programming and greedy heuristics to solve it.

Findings

Proposed algorithms outperform baseline methods in influence maximization.

Linear programming relaxation effectively approximates the NP-hard problem.

Heuristic approaches achieve balanced influence among products.

Abstract

The billboard advertisement has emerged as an effective out-of-home advertisement technique where the objective is to choose a limited number of slots to play some advertisement content (e.g., animation, video, etc.) with the hope that the content will be visible to a large number of travelers, and this will be helpful to earn more revenue. In this paper, we study a variant of the influential slot selection problem where the advertiser wants to promote multiple products. Formally, we call this problem the \textsc{Multi-Product Influence Maximization Problem for the Balanced Popularity} Problem. The input to our problem is a trajectory and a billboard database, as well as a budget for each product. The goal here is to choose a subset of slots for each product such that the aggregated influence of all the products gets maximized subject to the following two constraints: total selection cost for each product is less than or equal to the allocated budget for that product, and the difference between the influence for any two products is less than or equal to a given threshold. We show that the problem is NP-hard to solve optimally. We formulate this problem as a linear programming problem and use linear programming relaxation with randomized rounding. Further, we propose a greedy-based heuristic with balance correction to solve this problem. We conduct a number of experiments with real-world trajectory and billboard datasets, and the results are reported. From the reported results, we observe that the proposed solution approaches lead to more influence compared to many baseline methods.