基于泛化熵随机森林的雷达目标航迹起始方法

文章提出了一种基于泛化熵随机森林的雷达目标航迹起始方法，通过引入一种泛化熵可调参数，解决了传统随机森林算法对差异数据集泛化能力和适应性差且难以确定全局最优解的难题。首先利用雷达多周期回波数据构建随机森林算法决策模型，其次基于设计的泛化熵处理规则对实测数据进行分类处理得到所需测试样本，最后将测试样本输入完成模型训练的随机森林进行迭代寻优和自主决策以获取雷达目标的航迹起始结果。雷达实测数据验证表明，文章所提出的方法相比于传统随机森林方法具有更为优越的目标航迹起始性能。

Track initiation based on random forests with generalized entropy

A track initiation approach based on random forests with generalized entropy has been proposed in this paper. In the framework of generalized entropy, the generalization performance of decision tree in random forest is improved on various data sets by introducing an adjustable parameter. Furthermore, it can effectively over-come the problems of poor adaptability to data and local optimization in the process of tree building. In this work, the radar target track initiation problem has been transformed into sample classification with supervised learning. In the first stage, the random forest decision tree model has been constructed based on historical data. Then the measurement data are preprocessed using the generalized entropy adjustable simple rule method to get the test samples. Finally, the test samples will be input the trained random forest to get the target track initi-ation result. It is verified by radar data that the proposed method has better track initiation performance com-pared with the traditional random forest method.