Research On The Application Of Feature Selection Based On Rough Sets And Ant Colony Optimization Method

Feature selection has become the focus research in the field of data mining, machine learning, pattern recognition and so on.Feature selection uses a more stable set and appropriate precision characteristics to describe the original feature set.Feature selection research has focused on two aspects:one is for the search strategy of the subset and the other one is the performance evaluation feature subset. Therefore, the research on more effective feature selection algorithm, to obtain the better feature subset,to reduce the time complexity of the algorithm and to find the fast feature selection algorithm is still focus of the study of feature selection. According to the defects and deficiencies of the current algorithm, this paper proposed a new method for feature selection which combined the rough set method and ant colony optimization algorithm.The main contribution of this thesis includes:Firstly, the thesis reviews the theories and methods of rough set, include information expression system,the approximation and lower approximation, attribute reduction and core, attribute dependency, and the concepts of importance and the summary overview of the theoretical knowledge of the ant colony algorithm.Secondly, we have study and analysis the feature selection algorithms.In particular, feature selection algorithms based on rough sets (greedy method) and the ant colony optimization have been studied.Thirdly, by analyzing the advantages and disadvantages of the existing algorithms,the current shortcomings and deficiencies of methods have been found to propose a new method for feature selection which combined the rough set method and ant colony optimization algorithm.To improve the algorithm's performance, the core attribute as the start of the feature selection.In the transfer rules and the pheromone update strategy, this algorithm uses rough set dependency and attributes significance to guide the ants search process to improve the performance of the algorithm.In addition,the quality of classification based on rough set method and the length of the feature subset are used to measure the strengths and weaknesses of feature subset. By choosing a data set with certain number of data and attributes the proposed method is tested to compare with the feature selection method based on rough set and the feature selection method based on ant colony optimization.Testing and comparison results show that the proposed method is feasible and this method has obvious advantages in the indicators feature subset length and accuracy when the data set have core attributes.Finally, the paper research was summarized, and the future work prospected.