Ant Colony Algorithm And Its Applications In Data Mining

Natural computation has developed to be an important branch of computer science and is of great development prospect. There are many natural computation algorithms under research, for example Genetic Algorithm, Immune Algorithm, Neural Networks at early time and many new natural computation algorithms proposed in 1990s.The new natural computation algorithms include Ant colony algorithm, Quantum Computing, DNA Computing and Membrane Computing.Ant Colony Algorithm is proposed by the different behaviors of ant colony and has the characters of systematism, distribution, global convergence etc. According to the difference of ant colony behaviors, ant colony algorithm is classified into different types which are the ant colony optimization based on ant colony foraging behavior, the ant colony clustering algorithm based on ant nest cleaning behaviors and others. So far, the application fields of ant colony optimization have been extended from combination optimization to network routing, robot path planning, imagine processing etc and the ant colony clustering algorithm has been applied into the fields of data mining, data analysis, graph coloring problem etc.It is the research hotspot recently that the application of ant colony algorithm in data mining field, such as data classification and data clustering. But there are not enough relative research achievements. This paper is to analyze the ant colony algorithm in detail and change the running mechanism of ant colony algorithm to improve the application of algorithm in data mining.The main research working and achievements in the paper as follows:①. Expound the history and research hotspot of ant colony algorithm systematically;②. Analysis the configuration of parameters in ant colony algorithm in detail based on TSP; Get the relationship between the different parameters of ant colony algorithms through lots of numerical experiments; Analysis and get the relationship between configuration of parameters and algorithm efficient or performance; Proposed a new way in configuration of parameters.③. Analysis the characters of Ant-Miner in detail and proposed improved strategies on Ant-Miner aim at data classification rule mining. The improved strategies include Two-Step strategy used to expand the ants searching room and the improved pheromone updating mechanism which can improve the algorithm efficient and performance.④. Introduce different ant clustering algorithms based on ant colony foraging behavior and ant nest cleaning behavior.The analysis of configuration of parameters in this paper is benefit for the theoretical and experiment analysis of ant colony algorithm; The research on combinational of ant colony algorithm parameters is helpful to get more efficient and better performance ant colony algorithm; The improved strategies on Ant-Miner introduced in this paper is helpful for discovering better data classification algorithms.