有混合数据输入的自适应模糊神经推理系统

张宇献; 郭佳强; 钱小毅; 王建辉

doi:10.16383/j.aas.2018.c170698

有混合数据输入的自适应模糊神经推理系统

doi: 10.16383/j.aas.2018.c170698

1.
沈阳工业大学电气工程学院沈阳 110870
2.
沈阳工业大学信息科学与工程学院沈阳 110870
3.
东北大学信息科学与工程学院沈阳 110819

基金项目:

辽宁省教育厅项目 LQGD2017035

辽宁省自然科学基金 2015020064

国家自然科学基金 61102124

详细信息

作者简介:
郭佳强沈阳工业大学信息科学与工程学院硕士研究生.主要研究方向为智能控制, 复杂系统建模.E-mail:guo_dataworld@163.com

钱小毅沈阳工业大学电气工程学院博士研究生.主要研究方向为智能优化, 复杂机电装备的故障诊断.E-mail:qianxiaoyi123@163.com

王建辉博士, 东北大学信息科学与工程学院教授.主要研究方向为智能控制, 复杂系统建模, 康复机器人.E-mail:wangjianhui@ise.neu.ediu.cn

通讯作者:
张宇献沈阳工业大学电气工程学院副教授.2007年获得东北大学控制理论与控制工程专业博士学位.主要研究方向为智能控制, 复杂系统建模, 智能优化.本文通信作者.E-mail:yuxian524524@163.com

计量
- 文章访问数: 2067
- HTML全文浏览量: 290
- PDF下载量: 104
- 被引次数: 0
出版历程
- 收稿日期: 2017-12-11
- 录用日期: 2018-02-26
- 刊出日期: 2019-09-20

An Adaptive Network-based Fuzzy Inference System with Mixed Data Inputs

1.
School of Electrical Engineering, Shenyang University of Technology, Shenyang 110870
2.
School of Information Science and Engineering, Shenyang University of Technology, Shenyang 110870
3.
College of Information Science and Engineering, Northeastern University, Shenyang 110819

Funds:

Educational Commission of Liaoning Province LQGD2017035

Natural Science Foundation of Liaoning Province 2015020064

National Natural Science Foundation of China 61102124

More Information

Author Bio:
Master student at the School of Information Science and Engineering, Shenyang University of Technology. His research interest covers intelligent control and complex system modeling

Ph.D. candidate at the School of Electrical Engineering, Shenyang University of Technology. His research interest covers intelligent optimization and fault diagnosis for complex mechanical and electrical equipment

Ph.D. professor at the College of Information Science and Engineering, Northeastern University. Her research interest covers intelligent control, complex system modeling and rehabilitation robot

Corresponding author: ZHANG Yu-Xian Associate professor at the School of Electrical Engineering, Shenyang University of Technology. He received his Ph. D. degree from Northeastern University in 2007. His research interest covers intelligent control, complex system modeling and intelligent optimization. Corresponding author of this paper

摘要

摘要: 现有数据建模方法大多依赖于定量的数值信息，而对于数值与分类混合输入的数据建模问题往往根据分类变量组合建立多个子模型，当有多个分类变量输入时易出现子模型数据分布不均匀、训练耗时长等问题.针对上述问题，提出一种具有混合数据输入的自适应模糊神经推理系统模型，在自适应模糊推理系统的基础上，引入激励强度转移矩阵和结论影响矩阵，采用基于高氏距离的减法聚类辨识模型结构，通过混合学习算法训练模型参数，使数值与分类混合数据对模糊规则的前后件参数同时产生作用，共同影响模型输出.仿真实验分析了分类数据对模型规则后件的作用以及结构辨识算法对模糊规则数的影响，与其他几种混合数据建模方法对比表明本文所提出的模型具有较高的预测精度和计算效率.
- 自适应模糊推理系统 /
- 结构辨识 /
- 激励强度转移矩阵 /
- 后件影响矩阵 /
- 混合属性数据
Abstract: The available data modeling methods mostly depend on quantitative numerical information. But the data modeling with both numerical and categorical data input often has to build multiple sub-models on the basic of combination of categorical variables. It is likely to present unevenly data distribution of sub-models, time-consuming training process and other problems when the multiple categorical variables are input. For the above problems, an adaptive network-based fuzzy inference system with mixed data inputs is proposed. Based on the structure of the adaptive network-based fuzzy inference system, a firing-strength transform matrix and a consequent influence matrix are introduced. The subtractive clustering based on the Gaussian distance is adapted to identify structure of model, and a hybrid learning algorithm is used to train parameters of model. The numerical and categorical data play an important role on the antecedent and consequent parameters of fuzzy rules, and jointly affect the output of model. The simulation experiment analyzes the effect on categorical data to the consequent rules and structure identification to number of fuzzy rules. Comparing with others data modeling with mixed data inputs, the proposed model in this paper has higher prediction accuracy and computational efficiency.
- Adaptive network-based fuzzy inference system /
- structure identification /
- firing-strength transform matrix /
- consequent influence matrix /
- mixed attribute data
注释:

1) 本文责任编委刘艳军

HTML全文

图 1 MDI-ANFIS结构

Fig. 1 Structure of MDI-ANFIS

下载: 全尺寸图片幻灯片

图 2 样本平均规则后件输出

Fig. 2 Average consequent output of samples

下载: 全尺寸图片幻灯片

图 3 模型训练误差对比

Fig. 3 Comparison of model training error

下载: 全尺寸图片幻灯片

图 4 模型预测结果对比

Fig. 4 Comparison of model prediction

下载: 全尺寸图片幻灯片

图 5 聚类结果对比图

Fig. 5 Comparison of clustering results

下载: 全尺寸图片幻灯片

图 6 模型训练误差

Fig. 6 Model training error

下载: 全尺寸图片幻灯片

图 7 MDI-ANFIS模型预测对比

Fig. 7 Prediction results comparison of MDI-ANFIS

下载: 全尺寸图片幻灯片

表 1 MDI-ANFIS混合学习算法

Table 1 Hybrid learning algorithm of MDI-ANFIS

参数集	算法
P^c, Pⁱ	LSE
P^t	LSE
P^p	BP

下载: 导出CSV

表 2 两种算法的平均规则后件影响和误差

Table 2 Average consequent influences and errors of two algorithms

组号	样本点个数	平均规则后件值		预测误差
组号	样本点个数	C-ANFIS	MDI-ANFIS	C-ANFIS	MDI-ANFIS
1	200	19.659	17.735	2.040	1.519
2	200	18.905	36.270	1.690	1.463
3	200	21.323	27.297	2.980	1.881
4	400	34.202	66.760	3.230	1.604
5	400	14.050	39.905	2.330	2.145
6	400	16.385	35.070	2.510	2.002
7	500	18.901	17.804	3.680	2.194
8	600	21.659	30.857	2.290	2.395
9	600	16.299	22.267	2.800	2.242
10	600	18.426	34.818	3.730	2.187
平均值	410	19.981	32.878	2.728	1.963

下载: 导出CSV

表 3 结构辨识性能对比

Table 3 Performance comparison of structure identification

组号	样本点个数	规则数		预测误差
组号	样本点个数	SC	GDSC	SC	GDSC
1	100	50	13	0.452	0.356
2	100	32	14	0.575	0.466
3	200	37	21	0.517	0.709
4	200	25	14	0.908	0.613
5	300	40	18	0.586	0.690
6	300	34	16	0.661	0.705
7	400	31	16	0.642	0.459
8	400	32	14	0.630	0.747
9	500	30	13	0.788	0.836
10	506	30	14	0.726	0.827
平均值	300	34	15	0.648	0.641

下载: 导出CSV

表 4 UCI数据集模型误差对比

Table 4 Model error comparison on UCI dataset

数据集	样本个数	混合属性 (N, C)	预测误差						误差降低率
数据集	样本个数	混合属性 (N, C)	ANFIS	N-ANFIS	F-ANFIS	S-MLP	C-ANFIS	MDI-ANFIS	ANFIS	N-ANFIS	F-ANFIS	S-MLP	C-ANFIS
Abalone	4 177	7, 1	2.608	1.842	1.997	3.985	2.632	1.951	0.336	-0.056	0.023	1.04	0.349
Boston Housing	506	11, 2	0.779	0.631	0.657	7.096	0.824	0.638	0.221	-0.011	0.029	10.1	0.291
Auto MPG	398	4, 3	2.072	0.912	0.871	6.969	0.963	0.605	2.42	0.507	0.439	10.5	0.591
Servo	167	2, 2	1.012	0.060	0.051	3.119	0.362	0.025	39.4	1.40	1.04	123	13.4
TAE	151	1, 4	2.972	0.196	0.385	0.849	0.192	0.225	12.2	-0.128	0.711	2.77	-0.146
Zoo	101	1, 15	1.276	0.062	0.059	2.542	0.126	0.072	16.7	-0.138	-0.181	34.3	0.750
Heart Disease	303	6, 7	0.255	0.073	0.062	1.483	0.108	0.086	1.96	-0.151	-0.279	16.2	0.255
平均值	-	-	1.568	0.539	0.583	3.720	0.744	0.515	10.462	0.203	0.255	28.273	2.213

下载: 导出CSV

参考文献(22)

[1]	Alexander F J, Hoisie A, Szalay A. Big data. Computing in Science & Engineering, 2011, 13(6):10-13 http://d.old.wanfangdata.com.cn/Periodical/dlxtzdh201601001
[2]	Provost F, Fawcett T. Data science and its relationship to big data and data-driven decision making. Big Data, 2013, 1(1):51-59 https://www.ncbi.nlm.nih.gov/pubmed/27447038
[3]	Wu X D, Zhu X Q, Wu G Q, Ding W. Data mining with big data. IEEE Transactions on Knowledge and Data Engineering, 2014, 26(1):97-107 doi: 10.1109/TKDE.2013.109
[4]	陈晋音, 何辉豪.基于密度的聚类中心自动确定的混合属性数据聚类算法研究.自动化学报, 2015, 41(10):1798-1813 http://www.aas.net.cn/CN/abstract/abstract18754.shtml Chen Jin-Yin, He Hui-Hao. Research on density-based clustering algorithm for mixed data with determine cluster centers automatically. Acta Automatica Sinica, 2015, 41(10):1798-1813 http://www.aas.net.cn/CN/abstract/abstract18754.shtml
[5]	Jacobs R A, Jordan M I, Nowlan S J, Hinton G E. Adaptive mixtures of local experts. Neural Computation, 1991, 3(1):79-87 doi: 10.1162-neco.1991.3.1.79/
[6]	Lee K W, Lee T. Design of neural networks for multi-value regression. In:Proceedings of the 2001 International Joint Conference on Neural Networks. Washington DC, USA:IEEE, 2001. 93-98 http://d.old.wanfangdata.com.cn/OAPaper/oai_arXiv.org_0705.1309
[7]	Brouwer R K. A feed-forward network for input that is both categorical and quantitative. Neural Networks, 2002, 15(7):881-890 doi: 10.1016/S0893-6080(02)00090-4
[8]	Brouwer R K. A hybrid neural network for input that is both categorical and quantitative. International Journal of Intelligent Systems, 2004, 19(10):979-1001 doi: 10.1002/int.20032
[9]	Rey-del-Castillo P, Cardeñosa J. Fuzzy min-max neural networks for categorical data:application to missing data imputation. Neural Computing and Applications, 2012, 21(6):1349-1362 doi: 10.1007/s00521-011-0574-x
[10]	Hsu C C. Generalizing self-organizing map for categorical data. IEEE Transactions on Neural Networks, 2006, 17(2):294-304 doi: 10.1109/TNN.2005.863415
[11]	张宇献, 彭辉灯, 王建辉.基于异构值差度量的SOM混合属性数据聚类算法.仪器仪表学报, 2016, 37(11):2555-2562 doi: 10.3969/j.issn.0254-3087.2016.11.019 Zhang Yu-Xian, Peng Hui-Deng, Wang Jian-Hui. Self-organizing mapping clustering algorithm based on heterogeneous value difference metric for mixed attribute data. Chinese Journal of Scientific Instrument, 2016, 37(11):2555-2562 doi: 10.3969/j.issn.0254-3087.2016.11.019
[12]	Liu M, Dong M Y, Wu C. A new ANFIS for parameter prediction with numeric and categorical inputs. IEEE Transactions on Automation Science and Engineering, 2010, 7(3):645-653 doi: 10.1109/TASE.2010.2045499
[13]	Jang J S R. ANFIS:adaptive-network-based fuzzy inference system. IEEE Transactions on Systems, Man, and Cybernetics, 1993, 23(3):665-685 doi: 10.1109/21.256541
[14]	Abdelrahim E M, Yahagi T. A new transformed input-domain ANFIS for highly nonlinear system modeling and prediction. In:Proceedings of the 2001 Canadian Conference on Electrical and Computer Engineering. Toronto, Canada:IEEE, 2001. 655-660 https://ieeexplore.ieee.org/document/933761
[15]	Mar J, Lin F J. An ANFIS controller for the car-following collision prevention system. IEEE Transactions on Vehicular Technology, 2001, 50(4):1106-1113 doi: 10.1109/25.938584
[16]	Lima C A M, Coelho A L V, Von Zuben F J. Fuzzy systems design via ensembles of ANFIS. In:Proceedings of the 2002 IEEE International Conference on Fuzzy Systems. Honolulu, USA:IEEE, 2002. 506-511 https://ieeexplore.ieee.org/document/1005042
[17]	Paramasivam S, Arumugan R, Umamaheswari B, Vijayan S, Balamurugan S, Venkatesan G. Accurate rotor position estimation for switched reluctance motor using ANFIS. In:Proceedings of the 2001 Conference on Convergent Technologies for the Asia-Pacific Region. Bangalore, India:IEEE, 2003. 1493-1497 http://ieeexplore.ieee.org/xpls/icp.jsp?arnumber=1273168
[18]	Lih W C, Bukkapatnam S T S, Rao P, Chandrasekharan N, Komanduri R. Adaptive neuro-fuzzy inference system modeling of MRR and WIWNU in CMP process with sparse experimental data. IEEE Transactions on Automation Science and Engineering, 2008, 5(1):71-83 doi: 10.1109/TASE.2007.911683
[19]	Chiu S L. Fuzzy model identification based on cluster estimation. Journal of Intelligent and Fuzzy Systems:Applications in Engineering and Technology, 1994, 2(3):267-278 https://dl.acm.org/citation.cfm?id=2656640
[20]	Tuerhong G, Kim S B. Gower distance-based multivariate control charts for a mixture of continuous and categorical variables. Expert Systems with Applications, 2014, 41(4):1701-1707 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=efe4850a38e280a7baffd18ce342e4e3
[21]	曾珂, 张乃尧, 徐文立.线性T-S模糊系统作为通用逼近器的充分条件.自动化学报, 2001, 27(5):606-612 http://d.old.wanfangdata.com.cn/Periodical/zdhxb200105002 Zeng Ke, Zhang Nai-Yao, Xu Wen-Li. Sufficient condition for linear T-S fuzzy systems as universal approximators. Acta Automatica Sinica, 2001, 27(5):606-612 http://d.old.wanfangdata.com.cn/Periodical/zdhxb200105002
[22]	刘慧林, 冯汝鹏, 胡瑞栋, 刘春华.模糊系统作为通用逼近器的10年历程.控制与决策, 2004, 19(4):367-371 doi: 10.3321/j.issn:1001-0920.2004.04.002 Liu Hui-Lin, Feng Ru-Peng, Hu Rui-Dong, Liu Chun-Hua. Decennary development of fuzzy systems as universal approximators. Control and Decision, 2004, 19(4):367-371 doi: 10.3321/j.issn:1001-0920.2004.04.002