AccScience Publishing / IJOCTA / Volume 12 / Issue 1 / DOI: 10.11121/ijocta.2022.1084
RESEARCH ARTICLE

Multi-objective regression modeling for natural gas prediction  with ridge regression and CMARS

Ayşe Özmen1*
Show Less
1 Department of Mathematics, TED University, Turkey
2 Department of Mathematics and Statistics, University of Calgary, Canada
IJOCTA 2022, 12(1), 56–65; https://doi.org/10.11121/ijocta.2022.1084
Submitted: 17 February 2021 | Accepted: 23 November 2021 | Published: 1 January 2022
© 2022 by the Author(s). This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution -Noncommercial 4.0 International License (CC-by the license) ( https://creativecommons.org/licenses/by-nc/4.0/ )
Abstract

Residential customers are the main users generally need a great quantity of natural  gas in distribution systems, especially, in the wintry weather season since it is  particularly consumed for cooking and space heating. Hence, it ought to be noninterruptible. Since distribution systems have a restricted ability for supply,  reasonable planning and prediction through the whole year, especially in winter  seasons, have emerged as vital. The Ridge Regression (RR) is formulated mainly  to decrease collinearity results through shrinking the regression coefficients and  reducing the impact in the model of variables. Conic multivariate adaptive  regression splines ((C)MARS) model is constructed as an effective choice for  MARS by using inverse problems, statistical learning, and multi-objective  optimization theories. In this approach, the model complexity is penalized in the  structure of RR and it is constructed a relaxation by utilizing continuous  optimization, called Conic Quadratic Programming (CQP). In this study, CMARS  and RR are applied to obtain forecasts of residential natural gas demand for local  distribution companies (LDCs) that require short-term forecasts, and the model  performances are compared by using some criteria. Here, our analysis shows that  CMARS models outperform RR models. For one-day-ahead forecasts, CMARS  yields a MAPE of about 4.8%, while the same value under RR reaches 8.5%. As  the forecast horizon increases, it can be seen that the performance of the methods  becomes worse, and for a forecast one week ahead, the MAPE values for CMARS  and RR are 9.9% and 18.3%, respectively.

Keywords
Ridge Regression
(C)MARS
CQP
Interior Point
Prediction of Natural Gas Consumption
Conflict of interest
The authors declare they have no competing interests.
References

[1] Özmen, A., (2021). Sparse regression modeling for  short- and long‐term natural gas demand  prediction. Annals Operations Research, DOI:  https://doi.org/10.1007/s10479-021-04089-x

[2] Özmen, A., Yılmaz Y., & Weber, G.-W. (2018).  Natural Gas Consumption Forecasting Model for  Residential User. Energy Economics, 70, 357–381. 

[3] Aster, R.C., Borchers, B., & Thurber C.H. (2013).  Parameter Estimation and Inverse Problems.  Academic Press, Boston.

[4] James, G., Witten, D., Hastie, T., & Tibshirani, R.  (2013). An introduction of statistical learning,  Springer.

[5] Wheeler, D. (2009). Simultaneous Coefficient  Penalization and Model Selection in Geographically Weighted Regression: The Geographically  Weighted Lasso. Environment and Planning A, 39(3), 722-742.

[6] Friedman, J.H. (1991). Multivariate adaptive  regression splines. Annals of Statistics,19,1–67.

[7] Taylan, P., Weber, G.-W., & Beck, A. (2007). A new  approach to regression by generalized additive  models and continuous optimization for modern  applications in finance, science and technology,  Optimization. 56, 5-6, 675–698.

[8] Taylan, P., Weber, G. W., & Özkurt, F. Y. (2010). A  new approach to multivariate adaptive regression  splines by using Tikhonov regularization and  continuous optimization. Top, 18(2), 377-395.

[9] Weber, G.-W., Batmaz, İ., Köksal, G., Taylan, P., &  Yerlikaya-Özkurt, F. (2012). CMARS: a new  contribution to nonparametric regression with  multivariate adaptive regression splines supported  by continuous optimization. Inverse Problems in  Science and Engineering, 20(3), 371-400.

[10] Ayyıldız, E., Purutçuoğlu, V., & Weber, G. W.  (2018). Loop-based conic multivariate adaptive  regression splines is a novel method for advanced  construction of complex biological  networks. European Journal of Operational  Research, 270(3), 852-861.

[11] Roos, C., Terlaky, T., & Vial, J. (1997). Interior  Point Approach to Linear Optimization: Theory and  Algorithms, John Wiley & Sons, New York.

[12] Roos, C., Terlaky, T., & Vial, J. (2006). Interior  Point Methods for Linear Optimization. Springer  Science, Heidelberg/Boston.

[13] MOSEK; software available at  http://www.mosek.com.

[14] Tibshirani, R.(1996). Regression shrinkage and  selection via the lasso. Journal of the Royal  Statistical Soceity B, 267-288.

[15] Hastie, T., Tibshirani, R. & Friedman, J.H. (2001).  The Element of Statistical Learning. Springer Seriesin Statistics; New York, USA.

[16] Özmen, A., Kropat, E., & Weber, G.-W. (2014b).  Spline regression models for complex multi-modal  regulatory networks. Optimization Methods and  Software (OMS) 29(3), 515-534. 

[17] Kriner, M. (2007). Survival Analysis with  Multivariate Adaptive Regression Splines,  Dissertation, LMU Munchen: Faculty of  Mathematics, Computer Science and Statistics.

[18] Hastie, T. & Tibshirani, R.J. (1996). Discriminant  analysis by Gaussian mixtures, Journal of the Royal  Statistical Society (Ser. B) 58, 155–176.

[19] Leathwicka, J.R., Elith, J., & Hastie, T. (1993).  Comparative performance of generalized additive  models and multivariate adaptive regression splines  for statistical modelling of species distributions,  Ecological Modelling 199 (2), 188-196.

[20] Kuter, S., Weber, G.-W., Özmen, A.,& Akyürek, Z.  (2014). Modern applied mathematics for alternative  modelling of the atmospheric effects on  satellite images, Modeling, Optimization, Dynamics  and Bioeconomy I, Springer Proceedings in  Mathematics & Statistics, 73, 469-485.

[21] Gürbüz, B., & Sezer, M. (2020). Modified  operational matrix method for second-order  nonlinear ordinary differential equations with quadratic and cubic terms. An International Journal  of Optimization and Control: Theories &  Applications (IJOCTA), 10(2), 218-225.

[22] Salford Systems-Data Mining and Predictive  Analytics Software: MARS software available at  http://www.salfordsystems.com.

[23] Weber, G.-W., Çavuşoğlu, Z, & Özmen, A.  (2012).Predicting default probabilities in emerging  markets by conic generalized partial linear models  and their optimization. Optimization, 61(4), 443- 457.

[24] Özmen A., Weber, G.-W., & Kropat, E. (2012).  Robustification of conic generalized partial linear  models under polyhedral uncertainty. International  IFNA-ANS scientific Journal “Problems of  Nonlinear Analysis in Engineering Systems”, 2(38),  104-113.

[25] Özmen, A., Weber, G.-W., Çavuşoğlu, Z., &  Defterli, Ö. (2013). The new robust conic GPLM  method with an Application to Finance: prediction  of credit default. Journal of Global Optimization (JOGO), 56 (2), 233-249. 

[26] Özmen, A. (2016). Robust optimization of spline  models and complex regulatory networks - theory,  methods and application. Contribution to  Management Science, Springer.

[27] Özmen, A., Weber, G.-W., Batmaz, I., & Kropat, E.  (2011). RCMARS: Robustification of CMARS with  Different Scenarios under Polyhedral Uncertainty  Set. Communications in Nonlinear Science and  Numerical Simulation (CNSNS) 16 (12), 4780-4787.

Share
Back to top
An International Journal of Optimization and Control: Theories & Applications, Electronic ISSN: 2146-5703 Print ISSN: 2146-0957, Published by AccScience Publishing