摘要
提出基于改进蚁群优化算法和k近邻算法相结合的特征选择算法。利用k近邻分类器的分类精度和特征子集维数加权构造了综合适应度指标,利用改进蚁群算法的全局寻优和多次优解搜索能力实现特征子集搜索。针对传统蚁群算法在特征选择中可能含有冗余特征的问题,设计了局部细化搜索方式,使得特征选择结果不含冗余特征的同时提高了算法的收敛性。通过测试数据验证了算法的有效性和快速性后,将所提算法应用于10机39节点电力系统的安全评估问题,获得了良好的特征选择和稳定预测性能。
A novel embedded feature selection method based on improved ant colony optimization algorithm combined with the k-nearest neighbor（k-NN） classifier is proposed to tackle the feature selection problem.A weighted sum of the k-NN classification accuracy and the selected feature dimension form the fitness function.The improved ant colony optimization algorithm provides good global searching capability and multiple sub-optimal solutions.A local refinement searching scheme is designed to exclude the redundant features and improves the convergence rate.The feasibility and effectiveness of the proposed algorithm is first verified by a set of artificial test data and then applied to the power system security analysis problem.In the IEEE 10-unit-39-bus system,the proposed scheme obtains well-behaved security-related kernel features and provides good transient stability assessment performance.
出处
《电工技术学报》
EI
CSCD
北大核心
2010年第12期154-160,166,共8页
Transactions of China Electrotechnical Society
基金
国家自然科学基金资助项目(50407014)
关键词
特征选择
蚁群优化算法
k-近邻分类器
电力系统安全评估
Feature selection
ant colony optimization algorithm
k-nearest neighbor classifier
power system security analysis
