CAJ | 학술논문

针对目标属性识别的特点,建立了基于粗糙集(Rough Sets, RS)的数据分组处理(Group Method of Data Handling, GMDH)神经网络分类模型.该模型较好地解决了采用高维数据集训练神经网络效率低,神经网络结构规模较大的问题.同时为了提高高维数据集合的属性约简效率,改进了集合近似质量属性约简算法.最后,通过与BP(Back-Propagation, BP)神经网络分类能力的仿真对比,结果表明,基于粗糙集的数据分组处理神经网络分类模型分类能力优于BP神经网络模型,满足现代防空作战对目标属性识别的需求,基于快速求核和集合近似质量的属性约简算法快速有效.
침대목표속성식별적특점,건립료기우조조집(Rough Sets, RS)적수거분조처리(Group Method of Data Handling, GMDH)신경망락분류모형.해모형교호지해결료채용고유수거집훈련신경망락효솔저,신경망락결구규모교대적문제.동시위료제고고유수거집합적속성약간효솔,개진료집합근사질량속성약간산법.최후,통과여BP(Back-Propagation, BP)신경망락분류능력적방진대비,결과표명,기우조조집적수거분조처리신경망락분류모형분류능력우우BP신경망락모형,만족현대방공작전대목표속성식별적수구,기우쾌속구핵화집합근사질량적속성약간산법쾌속유효.
In the modern aerial defense fight, target attributes recognition is related to many factors, recognition process is complex, which calls for high time efficiency. A group method of data handling neural networks classification model is set up based on rough sets, aimed at characteristics of target attributes recognition. By using the model a lot of problems are solved, such as the low efficiency while high dimension data sets are used to train the neural networks and the neural networks configuration scale is great. Meanwhile, in order to boost the attributes reduction efficiency of high dimension data sets, the set approximate quality reduction algorithm is improved. Finally, in contrast with the simulation result of BP neural networks, the result shows that the classification quality of group method of data handling neural networks classification model based on rough sets is better than that of BP neural networks model, which satisfies the requirement for target attributes recognition in modern aerial defense fight, the attributes reduction algorithm based on speediness seeking core and set approximate quality is rapid and efficient.