关键词: 特征选择, 蜻蜓算法, 序列浮动后向选择, 群智能算法, 优化

Abstract: To properly apply the data generated in the manufacturing process,analyze and mine the key features and potential information in the data to help the enterprise improve efficiency and reduce costs,an improved dragonfly algorithm was proposed for feature selection.On the basis of inheriting the good convergence of dragonfly algorithm,the proposed algorithm introduced a local sequence floating backward selection mechanism,which could increase the global search ability of the algorithm and improve the accuracy of the algorithm.The algorithm was compared and analyzed with other popular feature selection algorithms on some standard complex datasets,and applied to real semiconductor production line for verification.The experimental results showed that the proposed algorithm had good global search ability and development potential in both standard data sets and real engineering problems.

Key words: feature selection, dragonfly algorithm, sequential backward floating selection, swarm intelligence algorithm, optimization