Toward Optimal Coupon Allocation in Social Networks: An Approximate Submodular Optimization Approach

TL;DR

This paper develops an approximate submodular optimization method to improve coupon allocation in social networks, maximizing influence spread under complex constraints, with theoretical guarantees close to optimal.

Contribution

It introduces a novel approximate algorithm for non-submodular utility functions in coupon allocation, achieving near-optimal approximation ratios under matroid and knapsack constraints.

Findings

Proposed an algorithm with approximation ratio approaching 1-1/e as epsilon approaches zero.

Applicable to a broad class of optimization problems modeled as approximate submodular maximization.

Achieved the best known approximation guarantees for this class of problems.

Abstract

CMO Council reports that 71\% of internet users in the U.S. were influenced by coupons and discounts when making their purchase decisions. It has also been shown that offering coupons to a small fraction of users (called seed users) may affect the purchase decisions of many other users in a social network. This motivates us to study the optimal coupon allocation problem, and our objective is to allocate coupons to a set of users so as to maximize the expected cascade. Different from existing studies on influence maximizaton (IM), our framework allows a general utility function and a more complex set of constraints. In particular, we formulate our problem as an approximate submodular maximization problem subject to matroid and knapsack constraints. Existing techniques relying on the submodularity of the utility function, such as greedy algorithm, can not work directly on a non-submodular function. We use $\epsilon$ to measure the difference between our function and its closest submodular function and propose a novel approximate algorithm with approximation ratio $\beta(\epsilon)$ with $\lim_{\epsilon\rightarrow 0}\beta(\epsilon)=1-1/e$. This is the best approximation guarantee for approximate submodular maximization subject to a partition matroid and knapsack constraints, our results apply to a broad range of optimization problems that can be formulated as an approximate submodular maximization problem.