GATeS: A Hybrid Algorithm Based on Genetic Algorithm and Tabu Search for the Direct Marketing Problem

Abstract

This paper deals with the problem of selecting a set of clients that will receive an offer of one or more products during a promotional campaign. Such campaigns are essential marketing tools to improve the economic profit of an enterprise, either by acquiring new customers or generating additional revenue from existing customers. In this research, a well-known mathematical model for the problem is used and extended with the cannibalism constraint which avoids some products being offered simultaneously to simulate competing products cannibalizing each other's market. To solve this problem, a hybrid heuristic is proposed, which uses a genetic algorithm (GA) as long-Term memory for a tabu search (TS). The main idea is not to use GA exclusively as an optimization procedure but also as a diversification strategy. In particular, GA elite solutions replace the TS's current solutions exploring in this way new areas in the search space. GA also receives the best TS solutions to maintain its population with high-quality solutions. Extensive computational experiments are performed on a set of existing benchmark test problems integrated with the restriction of cannibalism. A new set of instances with a high degree of difficulty is generated and are available to the research community through GitHub. The proposed method is compared with state-of-The-Art methods demonstrating better overall performance (sometimes more than 10 percentage points) and statistical significance.