Bionical Algorithms And Its Application On Expert Assignment Problem

Genetic algorithm (GA) and ant colony optimization (ACO) are two best popular bionical algorithms. GA is based on natural selection and evolution theory, and ACO simulates ants' behaviour of looking for food. This paper did much research on these two algorithms and proposed several new algorithms to improve their performance. Moreover this paper introduced them to a new discrete optimization area: expert assignment problem. The main works and innovative points are as follows:Two improved GAs for multimodal optimization were proposed. One is a novel niche genetic algorithm (NGA) with local search ability. The new algorithm adopted the mechanism of crossover and mutation in niche population instead of the whole population during late iterations. The results used in Shubert function showed its superiority. The other is a hybrid algorithm of HGA and Hopfield Neural Network (HNN). A group of solutions with variety were obtained using HGA firstly, and then the solutions were partitioned into some clusters whose centroids were as the initial value of each HNN, and HNNs were run to obtain all minima. It made use of the advantages of both HGA and HNN, and appeared excellent characteristic in optimal problems of multimodal function.A hybrid algorithm of GA and ACO were proposed. It added GA to ACO'every generation. Making use of GA's advantage of whole quick convergence, ACO'convergence speed was quickened. And GA's mutation mechanism improved the ability of ACO to avoid being premature.A new intelligent ACO for traveling salesman problem (TSP) was proposed. The new algorithm extracted the intrinsic characteristic rule of TSP and then injected it into the elite of ants, which improved the elite ant's capability to build a better solution and then made an improvement of ACO.Combining with the property of the expert assignment problem, this paper designed the improved genetic operators and ACO operations, and proposed methods of solving expert assignment problem using GA, ACO and hybrid algorithms of GA and ACO.