基于自适应遗传算法的雷达信号特征选择

Radar Signal Feature Selection Based on Self-Adaptive Genetic Algorithm

特征选择是机器学习领域的研究热点,是解决维数灾难的有效途径之一。采用一种新的自适应遗传算法和新的特征集评价准则(DFS)作为特征选择的方法。该遗传算法利用有限齐次马氏链进行建模,同时采用佳点集的方法,提高算法的准确度。该特征集评价准则通过计算特征集合中全部特征对于雷达信号分选的联合贡献来判断特征集合的优异性。仿真实验表明,该算法在雷达信号的特征选择的应用中,具有全局收敛性,有效地选择出最优特征集合,降低了特征维数,获得了更高的雷达信号分选正确率。

Feature selection is a hot topic in the field of machine learning,and it is one of the effective ways to solve the dimension disaster.A new adaptive genetic algorithm and a new feature set evaluation criterion(DFS)are used as the feature selection method.The genetic algorithm uses the finite homogeneous Markov chain for modeling,and the good point set is adopted to improve the accuracy of the algorithm.The feature set evaluation criterion determines the excellence of the feature set by calculating the joint contribution of all features to radar signal sorting in the feature set.The simulation experiment shows that the algorithm is globally convergent in the application of radar signal feature selection,the best feature set coulde be selected effectively,the feature dimension could be reduced,and a higher accuracy of radar signal sorting could be obtained.