Abstract
In this paper, we introduce an evolving system utilizing sparse weighted kernel least square as local models and online Gustafson-Kessel clustering algorithm for structure identification. Our proposed online clustering algorithm forms elliptical clusters with any orientation which leads to creating less but more complex shape clusters than spherical ones. Moreover, the clustering algorithm is able to determine number of required clusters by adding new clusters over time and to reduce the redundancy of model by merging similar clusters. Additionally, we propose weighted kernel recursive least square method with a new sparsification procedure based on instant prediction error. Also, we introduce an adaptive gradient-based rule for tuning kernel size. The sparsification procedure and adaptive kernel size improve the performance of kernel recursive least square, significantly. To illustrate our methodology, we apply the introduced model to online identification of a time varying and nonlinear system. Finally, to show the superiority of our approach in comparison to some known online approaches, two different time series are considered: Mackey–Glass as a benchmark and electrical load as a real-world time series.
Similar content being viewed by others
Abbreviations
- N :
-
Dictionary capacity
- n :
-
# of features/regressors
- M :
-
# of clusters/fuzzy rules
- m :
-
Fuzzification degree
- c i j :
-
ith stored data in jth local dictionary
- a i j :
-
ith consequence parameter of jth dictionary
- κ(.,.):
-
Kernel function
- x, y :
-
Input–output pair
- Q :
-
# of input–output pairs
- \(\hat{y}_{j}\) :
-
Estimated output by jth fuzzy rule
- \(\hat{y}\) :
-
Total estimated output
- Ψ j (x):
-
Fuzzy membership of x in jth fuzzy rule
- x k :
-
The kth input vector
- y k :
-
The kth output value
- μ k j :
-
Cluster center of jth cluster at step k
- A k j :
-
Norm inducing matrix of jth cluster at step k
- F k j :
-
Fuzzy covariance matrix of jth cluster at step k
- ρ j :
-
Cluster volume parameter of jth cluster
- N k j :
-
Membership sum of jth cluster at step k
- S j (x k):
-
Similarity value of kth input and jth cluster
- η 1 :
-
Thresholds value for creating new cluster
- sim pq :
-
Similarity value of pth and qth clusters
- η 2 :
-
Thresholds value for merging two clusters
- λ :
-
Regularization parameter
- ω(k):
-
Weight vector in feature space
- \(\varvec{\varphi }(k)\) :
-
Feature vector of x k
- G(k):
-
Kernel/gram matrix
- Y(k):
-
Output vector
- U(k):
-
Weight/membership matrix
- Φ(k):
-
Feature matrix
- a(k):
-
Consequence parameter at step k
- g(k):
-
Kernel vector
- \({\mathcal{D}}_{j}\) :
-
jth local dictionary
- Q, z j , r j :
-
Slack variables
- e k j :
-
Prediction error of jth fuzzy rule at step k
- \({\tilde{\mathbf{Q}}}_{j} \left( k \right)\) :
-
Reduced matrix of Q(k)
- C j (k):
-
The set of all stored input data in \({\mathcal{D}}_{j}\) at step k
- Y j (k):
-
The set of all stored output data in \({\mathcal{D}}_{j}\) at step k
- σ k j :
-
Kernel size of jth fuzzy rule at step k
- P k j :
-
Hessian matrix
- J k j :
-
Gradient vector
References
Angelov PP, Filev DP (2004) An approach to online identification of Takagi–Sugeno fuzzy models. IEEE Trans Syst Man Cybern B 34(1):484–498
Angelov P, Filev D (2005) Simpl_eTS: a simplified method for learning evolving Takagi-Sugeno fuzzy models. In: Proceedings of IEEE international conference on fuzzy systems, pp 1068–1073
Angelov P, Zhou X (2006) Evolving fuzzy systems from data streams in real-time. In: Proceedings of IEEE international symposium on evolving fuzzy systems, pp 29–35
Angelov P, Zhou X (2008) On line learning fuzzy rule-based system structure from data streams. In: Proceedings of IEEE international conference on fuzzy systems, pp 915–922
Angelov P, Giglio V, Guardiola C, Lughofer E, Luján JM (2006) An approach to model-based fault detection in industrial measurement systems with application to engine test benches. Meas Sci Technol 17(7):1809
Baruah RD, Angelov P (2011) Evolving fuzzy systems for data streams: a survey. Wiley Interdiscip Rev: Data Min Knowl Disc 1(6):461–476
Bezdek JC, Ehrlich R, Full W (1984) FCM: the fuzzy c-means clustering algorithm. Comput Geosci 10(2):191–203
Bittanti S, Piroddi L (1997) Nonlinear identification and control of a heat exchanger: a neural network approach. J Franklin Inst 334(1):135–153
Cortes C, Vapnik V (1995) Support-vector networks. Mach Learn 20(3):273–297
DaISy (2014) Database for the identification of systems. http://www.esat.kuleuven.be/sista/daisy/
Dovžan D, Škrjanc I (2011) Recursive clustering based on a Gustafson-Kessel algorithm. Evol Syst 2(1):15–24
Du H, Zhang N (2008) Application of evolving Takagi–Sugeno fuzzy model to nonlinear system identification. Appl Soft Comput 8(1):676–686
Engel Y, Mannor S, Meir R (2004) The kernel recursive least-squares algorithm. IEEE Trans Signal Process 52(8):2275–2285
Georgieva O, Filev D (2009) Gustafson-Kessel algorithm for evolving data stream clustering. In: Proceedings of the international conference on computer systems and technologies, pp 62:66
Gustafson DE, Kessel WC (1978) Fuzzy clustering with a fuzzy covariance matrix. In: Proceedings of IEEE conference on decision and control, pp 761–766
Kalhor A, Araabi BN, Lucas C (2010) An online predictor model as adaptive habitually linear and transiently nonlinear model. Evol Syst 1(1):29–41
Kalhor A, Araabi BN, Lucas C (2012) A new systematic design for habitually linear evolving TS fuzzy model. Expert Syst Appl 39(2):1725–1736
Kalhor A, Araabi BN, Lucas C (2013) Evolving Takagi–Sugeno fuzzy model based on switching to neighboring models. Appl Soft Comput 13(2):939–946
Kasabov NK, Song Q (2002) DENFIS: dynamic evolving neural-fuzzy inference system and its application for time-series prediction. IEEE Trans Fuzzy Syst 10(2):144–154
Komijani M, Lucas C, Araabi BN, Kalhor A (2012) Introducing evolving Takagi–Sugeno method based on local least squares support vector machine models. Evol Syst 3(2):81–93
Liu W, Park IM, Wang Y, Principe JC (2009) Extended kernel recursive least squares algorithm. IEEE Trans Signal Process 57(10):3801–3814
Liu W, Principe JC, Haykin S (2010) Kernel adaptive filtering: a comprehensive introduction. Wiley, Hoboken
Lughofer ED (2008) FLEXFIS: a robust incremental learning approach for evolving Takagi–Sugeno fuzzy models. IEEE Trans Fuzzy Syst 16(6):1393–1410
Lughofer E (2011) Evolving fuzzy systems—methodologies, advanced concepts and applications. Springer, Heidelberg
Lughofer E, Macián V, Guardiola C, Klement EP (2010) Data-driven design of Takagi–Sugeno fuzzy systems for predicting NOx emissions. In: Information processing and management of uncertainty in knowledge-based systems. Applications. Springer, Berlin, Heidelberg, pp 1–10
Muñoz A, Sánchez-Úbeda E, Cruz A, Marín J (2010) Short-term forecasting in power systems: a guided tour. In: Handbook of power systems II. Springer, Berlin, pp 129–160
Ngia LSH, Sjoberg J, Viberg M (1998) Adaptive neural nets filter using a recursive Levenberg-Marquardt search direction. In: Proceedings of the 32nd asilomar conference on signals, systems and computers, pp 697–701
Petersen KB, Pedersen MS (2008) The matrix cookbook. Technical University of Denmark
PJM (2014) PJM—historical metered load data. http://www.pjm.com/markets-and-operations/ops-analysis/historical-load-data.aspx
Sannen D, Nuttin M, Smith J, Tahir MA, Caleb-Solly P, Lughofer E, Eitzinger C (2008) An on-line interactive self-adaptive image classification framework. In: Computer vision systems. Springer, Berlin, pp 171–180
Scholkopf B, Smola AJ (2001) Learning with kernels: support vector machines, regularization, optimization, and beyond. MIT Press, Cambridge
Soleimani-B H, Lucas C, Araabi BN (2010) Recursive Gath-Geva clustering as a basis for evolving neuro-fuzzy modeling. Evol Syst 1(1):59–71
Soleimani-B H, Lucas C, Araabi B (2012) Fast evolving neuro-fuzzy model and its application in online classification and time series prediction. Pattern Anal Appl 15(3):279–288
Suykens JAK, Vandewalle J, De Moor B (2001) Optimal control by least squares support vector machines. Neural Netw 14(1):23–35
Takagi T, Sugeno M (1985) Fuzzy identification of systems and its applications to modeling and control. IEEE Trans Syst Man Cybern 15:116–132
Tipping ME (2001) Sparse Bayesian learning and the relevance vector machine. J Mach Learn Res 1:211–244
Van Vaerenbergh S, Via J, Santamaria I (2006) A sliding-window Kernel RLS algorithm and its application to nonlinear channel identification. In: Proceedings of IEEE international conference on acoustics, speech and signal processing, pp 789–792
Author information
Authors and Affiliations
Corresponding author
Appendix A
Appendix A
According to recursive Gauss–Newton algorithm, \({\mathbf{J}}_{j}^{k + 1} \, = \, - \frac{{\partial e^{k + 1} }}{{\partial \sigma_{j}^{k} }}\) can be written as following:
where \(\frac{{\partial {\mathbf{g}}_{j} \left( {k + 1} \right)}}{{\partial \sigma_{j}^{k} }}\) equals to:
So by considering all these, we can conclude (47).
Rights and permissions
About this article
Cite this article
Shafieezadeh-Abadeh, S., Kalhor, A. Evolving Takagi–Sugeno model based on online Gustafson-Kessel algorithm and kernel recursive least square method. Evolving Systems 7, 1–14 (2016). https://doi.org/10.1007/s12530-015-9129-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s12530-015-9129-1