基于粒子群优化算法和相关性分析的特征子集选择

Feature Subset Selection Based on Particle Swarm Optimization Algorithm and Relevance Analysis

特征选择是模式识别与数据挖掘等领域的重要问题之一。针对此问题,提出了基于离散粒子群和相关性分析的特征子集选择算法,算法中采用过滤模式的特征选择方法,通过分析网络入侵数据中所有特征之间的相关性,利用离散粒子群算法在所有特征的空间里优化搜索,自动选择有效的特征子集以降低数据维度。1999KDD Cup Data中IDS数据集的实验结果表明了提出算法的有效性。

Feature selection is one of the important problems in the pattern recognition and data mining areas. The new feature subset selection method based on discrete binary version of particle swarm optimization （PSO） algorithm and relevance analysis is proposed. This new method employs the filter mode feature selection algorithm, which focuses on the correlation among the features of the network traffic data and employs the discrete particle swarm algorithm to find an optimized feature set. Experiments in 1999 KDD Cup Data confirm the effectiveness of the proposed strategy.