Aircraft type prediction based on approximate rough set resolution classification algorithm
-
摘要:
传统的C4.5分类决策树作为数据分类算法具有计算简单、准确率高的优势,由于飞机具有参数多和数据量大的因素,C4.5算法需要对连续属性值进行多次顺序扫描,分类时间效率较低。针对此问题,提出近似粗糙集和决策分辨力分类算法,利用粗糙集近似度来判断属性划分样本数据能力,并将其代入到决策分辨力算法中,以决策分辨力最大的属性作为分裂特征建立分类决策树。算法在保证分类决策准确率的同时,提高计算效率并减少过拟合问题的产生。通过对UCI(University of California, Irvine)数据集上多组数据样本的对比实验分析,验证了本文提出PSRP(rough set and resolving power)的算法在保证相同准确率的情况下,平均计算时间效率提升约10%,可靠性提升2%。
Abstract:As a traditional data classification algorithm, C4.5 Decision Tree has the advantages of simple calculation and high accuracy. Due to the fact that aircraft has many parameters and large amount of data, C4.5 Decision Tree requires multiple sequential scanning of continuous attribute values, but the classification time efficiency is low. To solve this problem, an approximate rough set resolution classification algorithm was proposed. Rough set approximation was used to determine the ability of attributes to divide sample data. According to the probability acquired, the attribute with the largest resolution was used as the split character to establish the classification decision tree. The algorithm improved the computational efficiency and reduced the generation of over-fitting problem while ensuring the accuracy of classification decision. Through comparative experimental analysis of multiple sets of data samples on UCI (University of California, Irvine) databases, it proved that the PSRP (rough set and resolving power) algorithm improved the average computational time efficiency by about 10% and reliability by 2% while ensuring the same accuracy.
-
表 1 数据集描述
Table 1. Data set description
序号 数据集名称 样本个数 条件属性 决策类 1 heart failure 299 12 2 2 hcvdat 615 12 5 3 user model 258 5 4 4 liver patient 583 10 2 5 ionosphere 351 33 2 6 exasens 399 3 4 7 transfusion 748 4 2 8 wholesale customer 440 7 2 9 messidor 1151 19 2 10 spy plane finder 597 31 2 表 2 数据集分类评估指标对比
Table 2. Comparison of evaluation indexes for data set classification
数据集名称 F1 时间效率 树形复杂度 C4.5 RS-CART RSRP C4.5 RS-CART RSRP C4.5 RS-CART RSRP heart failure 0.7997 0.7715 0.8152 0.2829 0.1862 0.2421 6.78 4.42 3.98 hcvdat 0.9084 0.8616 0.8925 1.8430 1.0443 1.0243 7.90 4.82 4.00 user model 0.8954 0.8292 0.9065 0.2044 0.1767 0.1913 5.56 4.54 4.24 liver patient 0.6693 0.6942 0.7020 1.6517 0.8661 0.6875 44.09 6.44 6.39 ionosphere 0.8823 0.8920 0.8964 1.9100 1.0584 1.1292 4.52 3.64 3.46 exasens 0.5188 0.5214 0.5416 0.0789 0.0662 0.0734 8.02 6.12 6.54 transfusion 0.7591 0.7775 0.7695 0.2505 0.1905 0.1736 17.54 7.96 5.02 wholesale customer 0.8895 0.8996 0.9065 0.8743 0.6317 0.6417 8.62 5.20 4.40 messidor 0.5924 0.5551 0.6410 39.2418 7.7538 6.8217 104.2 10.24 6.6 平均值 0.7683 0.7558 0.7857 5.1486 1.3304 1.2205 23.0256 5.9311 4.9589 表 3 飞机类型预测评估指标对比
Table 3. Comparison of aircraft type prediction evaluation indexes
评价指标 C4.5 RS-CART RSRP F1 0.8710 0.8655 0.8744 时间效率 6.0691 5.5738 4.7178 树形复杂度 7.64 4.58 4.94 -
[1] COLUCCIA A,PARISI G,FASCISTA A. Detection and classification of multirotor drones in radar sensor networks: a review[J]. Sensors,2020,20(15): 1-22. doi: 10.1109/JSEN.2020.3003972 [2] LEI P,WANG J,GUO P,et al. Automatic classification of radar targets with micro-motions using entropy segmentation and time-frequency features[J]. Electron Commun,2011,65(10): 806-813. [3] ZHOU G Q,ZHANG R T,ZHANG D J. Manifold learning co-location decision tree for remotely sensed imagery classification[J]. Remote Sens,2016,8(855): 1-31. [4] 同亚龙,王彤,代保全,等. 一种抗密集欺骗式干扰的机载雷达动目标检测方法[J]. 电子与信息学报,2015,37(3): 658-664. doi: 10.11999/JEIT140679TONG Yalong,WANG Tong,DAI Baoquan,et al. A moving target detection method to suppress dense deception jamming for airborne radar[J]. Journal of Electronics & Information Technology,2015,37(3): 658-664. (in Chinese) doi: 10.11999/JEIT140679 [5] 王福友,罗钉,刘宏伟. 低分辨机载雷达飞机目标分类识别技术研究[J]. 雷达学报,2014,3(4): 444-449.WANG Fuyou,LUO Ding,LIU Hongwei. Low-resolution airborne radar aircraft target classification[J]. Journal of Radars,2014,3(4): 444-449. (in Chinese) [6] 雷涛,旷生玉,杨玲. 基于深度学习的雷达辐射源型号识别方法[J]. 电子信息对抗技术,2019,34(4): 29-34. doi: 10.3969/j.issn.1674-2230.2019.04.006LEI Tao,KUANG Shengyu,YANG Ling. Radar emitter type identification algorithm based on deep learning[J]. Electronic Information Warfare Technology,2019,34(4): 29-34. (in Chinese) doi: 10.3969/j.issn.1674-2230.2019.04.006 [7] 欧阳瑞麒,雍杨,王兵学. 卷积神经网络在飞机类型识别中的应用[J]. 兵工自动化,2017,36(12): 71-75.OUYANG Ruilin,YONG Yang,WANG Bingxue. Application of convolution neural network in aircraft type recognition[J]. Ordnance Industry Automation,2017,36(12): 71-75. (in Chinese) [8] 刘相云,龚志辉,金飞,等. 结合显著图和深度学习的遥感影像飞机目标识别[J]. 测绘通报,2019(3): 27-31.LIU Xiangyun,GONG Zhihui,JIN Fei,et al. Aircraft target recognition in remote sensing image combined saliency map with deep learning[J]. Bulletin of Surveying and Mapping,2019(3): 27-31. (in Chinese) [9] ZHANG K,ZHAN J M,WU W Z. Novel fuzzy rough set models and corresponding applications to multi- criteria decision-making[J]. Fuzzy Sets and Systems,2020(383): 92-126. [10] YAO Y Y,SHE Y H. Rough set models in multigranulation spaces[J]. Information Sciences,2016(327): 40-56. [11] 王越,万洪. 一种新的应用变精度粗糙集的决策树构造方法[J]. 重庆理工大学学报(自然科学),2013,27(11): 58-64.WANG Yue,WAN Hong. A new method for constructing decision tree based on variable precision rough set[J]. Journal of Chongqing University of Technology (Natural Science),2013,27(11): 58-64. (in Chinese) [12] 刘军. 基于叶枝比率的决策树构建算法[J]. 信息网络安全,2013(2): 9-12.LIU Jun. Algorithm of constructing decision tree based on the leaf and branch ratio[J]. Netinfo Security,2013(2): 9-12. (in Chinese) [13] JIANG J F,ZHU X Y,HAN G J,et al. A dynamic trust evaluation and update mechanism based on C4.5 decision tree in underwater wireless sensor networks[J]. IEEE Transactions on Vehicular Technology,2020,69(8): 31-40. [14] 尚朝轩,王品,韩壮志,等. 基于类决策树分类的特征层融合识别算法[J]. 控制与决策,2016,31(6): 1009-1014.SHANG Chaoxuan,WANG Pin,HAN Zhuangzhi,et al. Feature-level fusion recognition algorithm based on analogy decision tree classification[J]. Control and Decision,2016,31(6): 1009-1014. (in Chinese) [15] XIAO F Y. Multi-sensor data fusion based on the belief divergence measure of evidences and the belief entropy[J]. Information Fusion,2019(46): 23-32. [16] 陈泽华,张裕,谢刚. 不一致决策表规则获取的粒计算方法[J]. 控制与决策,2015,30(4): 709-714.CHEN Zehua,ZHANG Yu,XIE Gang. GrC method of rule acquisition for inconsistent decision table[J]. Control and Decision,2015,30(4): 709-714. (in Chinese) [17] LI J H,REN Y,MEI C L,et al. A comparative study of multigranulation rough sets and concept lattices via rule acquisition[J]. Knowledge-Based Systems,2016,91: 152-164. doi: 10.1016/j.knosys.2015.07.024 [18] 孙长嵩,董西国,张健沛. 一个基于粗糙集和决策树的最简分类规则集生成算法[J]. 哈尔滨工程大学学报,2002,23(5): 87-91.SUN Changsong,DONG Xiguo,ZHANG Jianpei. Algorithm based on rough set and decision tree to gain minimal classing rule set[J]. Journal of Harbin Engineering University,2002,23(5): 87-91. (in Chinese) [19] 王礼沅,张恒喜,徐浩军. 基于粗糙集的空战效能多指标综合评估模型[J]. 航空学报,2008,29(4): 880-885. doi: 10.3321/j.issn:1000-6893.2008.04.018WANG Liruan,ZHANG Hengxi,XU Haojun. Multi-index synthesize evaluation model based on rough set theory for air combat efficiency[J]. Acta Aeronautica et Astronautica Sinica,2008,29(4): 880-885. (in Chinese) doi: 10.3321/j.issn:1000-6893.2008.04.018 [20] JIAN L R, LI M Y. A variable precision fuzzy rough set approach to a fuzzy-rough decision table[J]. Science and Engineering Research Center, 2016: 250-254. [21] 张清华,薛玉斌,王国胤. 粗糙集的最优近似集[J]. 软件学报,2016,27(2): 295-308.ZHANG Qinghua,XUE Yubin,WANG Guoyin. Optimal approximation sets of rough sets[J]. Journal of Software,2016,27(2): 295-308. (in Chinese) [22] ZHANG X,MEI C L,CHEN D G,et al. Feature selection in mixed data: a method using a novel fuzzy rough set-based information entropy[J]. Pattern Recognition,2016(56): 1-15. [23] ZHAN J M,ALI M I,MEHMOOD N. On a novel uncertain soft set model: Z-soft fuzzy rough set model and corresponding decision making methods[J]. Applied Soft Computing,2017(56): 446-457. [24] ZHAN J M,LIU Q,HERAWAN T. A novel soft rough set: soft rough hemirings and corresponding multicriteria group decision making[J]. Applied Soft Computing,2017(54): 393-402. [25] 张铁柱,蒋宏. 机载雷达和红外数据融合的智能目标识别[J]. 红外与激光工程,2010,39(4): 756-760. doi: 10.3969/j.issn.1007-2276.2010.04.037ZHANG Tiezhu,JIANG Hong. Intelligent target recognitionbased on the data fusion of radar and infrared imaging sensor[J]. Infrared and Laser Engineering,2010,39(4): 756-760. (in Chinese) doi: 10.3969/j.issn.1007-2276.2010.04.037 [26] 李雪岩,李学伟,蒋君. 基于知识粒度特征的多目标粗糙集属性约简算法[J]. 控制与决策,2021,36(1): 196-205.LI Xueyan,LI Xuewei,JIANG Jun. Multi objective rough set attribute reduction algorithm based on the characteristics of knowledge granularity[J]. Control and Decision,2021,36(1): 196-205. (in Chinese) [27] BINH T P,ABOLFAZL J,INDRA P. A novel hybrid support vector machine with decision tree for data classification[J]. International Journal of Advanced and Applied Sciences,2017,4(9): 138-143. doi: 10.21833/ijaas.2017.09.019 [28] 庞梦洋,索中英,郑万泽,等. 基于RS-CART决策树的航空发动机小样本故障诊断[J]. 航空动力学报,2020,35(7): 1559-1568.PANG Mengyang,SUO Zhongying,ZHENG Wanze,et al. Small sample fault diagnosis of aeroengine based on RS-CART decision tree[J]. Journal of Aerospace Power,2020,35(7): 1559-1568. (in Chinese)