Research And Application Of Data Classification Means Based Improving Ant Colony Algorithm

Marco Dorigo and other scholars have proposed an algorithm of ant colony optimization, the algorithm is simulated the intelligent behavior of ant colony. After the neural networks, genetic algorithms, it solving the combinatorial optimization problem, assignment problem, scheduling problem also has a good results in this optimization algorithm. Experiments show that ant colony algorithm has so good performance, making it in the data mining algorithm for data classification applications also get good results, but the ant colony algorithm is always so easy to convergence to the shorter path. The work includes as follow:First, through in-depth study on the basic principles of ant colony algorithm and summarize recent progress, analyzing the advantages and disadvantages, we improve the strategy of algorithm, and proposes an algorithm of NIAS. It's applied to the traveling salesman problem (TSP). Search path through the intersection of the ant colony detection and adjust the path selection strategies aimed at improving the ability of ant colony to solve the optimal solution. And the intersection will change the pheromone update strategy. Experiments show that the improved algorithm can improve the performance of solving the TSP problem.Second, combining with the ACO, there are two ways of improving the Ant-Miner algorithm to classification rules from data:comparing the quality of each rule term by the random combination, the algorithm prune rules to make sure that the best rule term will be add to the discovered rule list. In order to avoid the premature, the pheromone will be enhanced the rule terms that not in the discovered rule list. Experiments showed that compared with basic ant colony classification algorithm (Ant-Miner), however the improved algorithm takes more times to get rules, but the result is better than the basic algorithm of classification.