Parallel Ant Colony Optimization Algorithms for Time Series Segmentation on a Multi-core Processor

This paper proposes four novel parallelization methods of a modified Ant Colony Optimization algorithm. The parallelization methods are aiming at finding the optimal segmentation scheme of time series with a low execution time. The series is decomposed into different sub-series firstly, and then each sub-series can be solved by colonies independently, finally merge the solutions of each colony to obtain the full segmentation scheme. According to the synchronization of individuals and colonies, we design four types of dual parallel models, and implement the parallel versions by using OpenMP library on a computing platform with a multi-core processor for time series segmentation. Experiment results suggest that the parallel algorithms can greatly shorten the execution time without reducing the quality of the final solution.