Creativity-Oriented Optimization Algorithm And Its Applications

The advent of the era of big data brings more opportunities for the development of science and technology, while it also lets us faces more challenges. Along with the expansion of the data scale, the complexity of the problems is increasing rapidly. Existing algorithms could hardly meet the coming demands. Recent years, nature-inspired algorithms have been gradually applied into practical problems. These algorithms are inspired by the laws or phenomena of nature and most of them have a certain adaptive, self-learning ability and inherent parallelism. The above features make these algorithms suitable for the complex problems, namely high-dimensional, highly nonlinear and random problems. Nature-inspired algorithms have broad application prospects in the age of big data. It’s meaningful and challengeable to continually exploring new high-performance nature-inspired algorithms.Creative thinking, which plays an essential role in the progress of human society, has an outstanding problem-solving ability. Inspired by the creative thinking process, this thesis presents a novel creativity-oriented optimization model (COOM). Compared with other nature-inspired algorithms, COOM is more intelligent and has a better parallel property. Based on this model, we have done the following works:(1) A creativity-oriented optimization algorithm (COOA) for the continuous optimization problems is proposed. First, the properties of COOA are discussed, including its convergence and parallelism. And then,28CEC-2013real-parameter benchmark functions are used to test the effectiveness and parallelism of the proposed approach. In addition, the parameters of COOA are analyzed in detail.(2) A binary creativity-oriented optimization algorithm (BCOOA) for the combination optimization problems is proposed. The convergence of BCOOA is analyzed theoretically at first. Then, it’s applied to the spectrum allocation problem in cognitive radio networks. Compared with some similar algorithms, BCOOA is more effective and has a good parallel property.(3) A classifier based on COOA is proposed. Simulation results on12UCI datasets are compared with8commonly used classification algorithms. It shows that, the classifier based on COOA has an appealing performance.