Multi-Colony Ant Colony Optimization Based on Generalized Jaccard Similarity Recommendation Strategy

Abstract

Ant Colony Optimization has achieved good results in solving Traveling Salesman Problem (TSP), it has a tendency to fall into local optima and the convergence speed is limited. To address this problem, multi-colony ant colony optimization based on the generalized Jaccard similarity recommendation strategy (JCACO) is proposed. Firstly, two classical ant populations, Ant Colony System and Max-Min Ant System are selected to form heterogeneous multi-colony. Secondly, attribute-based collaborative filtering recommendation mechanism is proposed to balance the diversity and convergence of the algorithm, three strategies have been implemented under this recommendation mechanism: The attribute cross-learning strategy is used to highlight the effect of excellent attributes and improve the attribute comprehensive performance; According to the diversity results of the population measured by information entropy, the attribute recommendation learning strategy is used to enrich the diversity of the population adaptively; The pheromone reward strategy is implemented on the public path to accelerate the convergence speed; Among which, according to the generalized Jaccard similarity coefficient, the most suitable communication object is recommended in order to achieve the best learning efficiency. Finally, when the algorithm stagnates, the elite reverse learning mechanism is used to jump out of the local optimum. Experimental results show that JCACO has good performance and high stability in TSP instances, especially in large-scale TSP instances.