Research on the Application of Feature Selection Based on Rough Sets and Ant Colony Optimization Method
|Course||Applied Computer Technology|
|Keywords||Rough Set Ant Colony Optimization Feature selection Attribute dependency Attribute importance|
Feature selection has become data mining, machine learning, pattern recognition, and other areas of research focus. Feature selection more stable feature set represents a collection of primitive features with appropriate accuracy. Feature selection studies mainly focused on two aspects, the Search feature subset search strategy, feature subset performance evaluation. Therefore, the study of more effective feature selection algorithm to get better feature subset, reduce the time complexity of the algorithm and to seek fast feature selection algorithm is still the focus of the study of feature selection. Based on rough set-based feature selection and optimization method based on ant colony characteristics select the algorithm two methods have advantages and drawbacks, and proposed a rough set method and ant colony optimization method combining feature selection algorithm. Expression of its main tasks include the following aspects: First, a brief introduction to the knowledge of the rough set theory and ant colony optimization algorithm, including information systems, on the Upper and Lower Approximation attribute reduction and nuclear properties dependence and an important degree of Overview summary of the concepts and theoretical knowledge of the ant colony algorithm. Secondly, the characteristics of the selection algorithm outline analysis. Depth study focuses on the degree of feature selection algorithm (greedy Act) based on rough sets and selection algorithm is based on the characteristics of the ant colony optimization method. Again, to analyze the characteristics of rough set-based feature selection method based on ant colony optimization algorithm has advantages and shortcomings on the basis of this paper, a set of characteristics of the ant colony optimization method based on rough selection algorithm. The proposed algorithm by introducing rough set relative nuclear attribute as the starting point of the feature selection, in order to improve the accuracy of the algorithm; transfer rules and pheromone update strategy, the introduction of a rough set attribute dependence and attribute importance for guiding ants the search process, in order to improve the performance of the algorithm; In addition, the rough set theory classification accuracy and feature subset length two parameters used in the evaluation function to measure the pros and cons of the feature subset; choose a different number of data and attributes The number of data sets of the proposed method was tested with both based on rough sets feature selection methods and feature selection algorithm based on ant colony optimization method compared experiment. Test and compare the experimental results show that the proposed method is feasible, and has obvious advantages in nuclear attribute data set on the length of the feature subset and the accuracy of two indicators. Finally, the thesis work are summarized and proposed further work outlook.