Pattern Recognition Letters 138 (2020) 8–15 Contents lists available at ScienceDirect Pattern Recognition Letters journal homepage: www.elsevier.com/locate/patrec A PSO-based algorithm for mining association rules using a guided exploration strategy  Gretel Bernal Baró a , José Francisco Martínez-Trinidad b , Rosa María Valdovinos Rosas a,∗ , Jesús A. Carrasco Ochoa b , Ansel Y. Rodríguez González c , Manuel S. Lazo Cortés d a Universidad Autónoma del Estado de México, Faculty of Engineering, Instituto Literario 100 Centro, Toluca de Lerdo 50000, México b Department of Computer Sciences at Institute of Astrophysics Optics and Electronics (INAOE), Luis Enrique Erro 1 Tonantzintla, Puebla 72840, México c Tepic Technology Transfer Unit, Center for Scientific Research and Higher Education at Ensenada (CICESE-UT3), Andador 10, Tepic, Nayarit 63173, México d TecNM/Instituto Tecnológico de Tlalnepantla, Av. Instituto Tecnológico s/n, La Comunidad, Tlalnepantla de Baz 54070, México a r t i c l e i n f o Article history: Received 10 October 2019 Revised 28 April 2020 Accepted 4 May 2020 MSC: 41A05 41A10 65D05 65D17 Keywords: PSO Algorithm Association Rules Metaheuristic algorithm a b s t r a c t Association rule mining is one of the most important and active research areas in data mining. In the lit- erature, several association rule miners have been proposed; among them, those based on particle swarm optimization (PSO) have reported the best results. However, these algorithms tend to prematurely fall into local solutions, avoiding a wide exploration that could produce even better results. In this paper, an algo- rithm based on PSO, called PSO-GES, for mining association rules using a Guided Exploration Strategy is introduced. Our experiments, over real-world transactional databases, show that our proposed algorithm mines better quality association rules than the most recent PSO-based algorithms for mining association rules of the state of the art. © 2020 Published by Elsevier B.V. 1. Introduction Association rule mining (ARM) is one of the most used tech- niques in data mining. The main purpose of association rule min- ing is to discover interesting associations between items in large transactional databases [1]. The ARM problem is defined as fol- lows: let T be the a set of M transactions T = t 1 , t 2 , . . . , t M , and I = {i 1 , i 2 , . . . , i n } be the set of all different items in T. An item- set X is a subset of items, i.e., X⊆I. The support of an itemset X in T, denoted as Sup (X), is the fraction of transactions in T that contain X. An itemset X is frequent if its support is no less than MinSup, where MinSup is a threshold given by the user. An associ- ation rule (AR) is an expression of the form X → Y, where X ∪ Y is a frequent itemset and X ∩ Y = ∅. X is called the antecedent and Y is called the consequent of the rule. Given a transactional database T, the ARM problem consists in finding all the association rules (ARs) that satisfy predefined minimum support and confidence thresh- olds. The support of an AR is the support of the union of its an-  Editor: Prof. S. Sarkar ∗ Corresponding author. E-mail address: li_rmvr@hotmail.com (R.M. Valdovinos Rosas). tecedent and its consequent, and the confidence of an AR is the support of the rule divided by the support of the antecedent of the rule. Algorithms for generating interesting ARs from frequent item- sets, have been reported in the literature, for example [6,19]. These algorithms mine all possible ARs, by first mining all the frequent itemsets through algorithms based on Apriori and FP-growth, and then the association rules are mined using the algorithm for min- ing ARs proposing in Agrawal et al. [1], therefore, they usually pro- duce a large number of ARs and selecting among them, a subset of good rules is a challenging and expensive task. For this reason, bio-inspired heuristics such as Ant Colony Optimization (ACO) [8], Particle Swarm Optimization (PSO) [2,11,18] and Genetic Algorithms (GA) [13] have been proposed for mining a good subset of ARs. From these works, the best results have been reported by those based on particle swarm optimization (PSO). However, algorithms based on PSO have the problem of prematurely falling into lo- cal solutions (rules) with low quality [4,14,20,21]. In this paper, a PSO-based algorithm named PSO-Guided Exploration Strategy (PSO-GES) for mining ARs is proposed, the main contribution is to include a Guided Exploration Strategy in order to generate high- quality rules. In addition, a new fitness function is introduced for https://doi.org/10.1016/j.patrec.2020.05.006 0167-8655/© 2020 Published by Elsevier B.V.