Abstract
Symbiotic organisms search (SOS) is a new metaheuristic optimization algorithm proposed by Cheng and Prayogo (Comput Struct 139:98–112, 2014). In this paper, SOS has been applied to determine the functional forms of different time series which are used to predict the time series. There are some previous attempts by researchers where genetic algorithm has been used to find the functional form of a time series. Here, we explore this new algorithm in time series analysis. SOS mimics the symbiotic relationships among organisms in the ecosystem. Improvement in SOS in parasitism phase has been proposed here. Also, several types of time series have been tested to compare the performance of the original SOS with its improved version and with already well-established artificial neural network (ANN) in the field of time series forecasting.
Similar content being viewed by others
References
Cheng M-Y, Prayogo D (2014) Symbiotic organisms search: a new metaheuristic optimization algorithm. Comput Struct 139:98–112
Szpiro GG (1997) Forecasting chaotic time series with genetic algorithms. Phys Rev E 55(3):2557–2568
Raicharoen T, Lursinsap C, Sanguanbhoki P (2003) Application of critical support vector machine to time series prediction. In: Proceedings of the 2003 international symposium on circuits and systems. ISCAS ’03, vol 5, pp 25–28, 741–744
Zhang GP (2003) Time series forecasting using a hybrid ARIMA and neural network model. Neurocomputing 50:159–175
Khashei M, Bijari M (2011) A novel hybridization of artificial neural networks and ARIMA models for time series forecasting. Appl Soft Comput 11:2664–2675
Yule GU (1927) On the method of investigating periodicities in disturbed series, with special reference to Wolfer’s sunspot numbers. Philos Trans R Soc Lond Ser A 226:267–298
Box GEP, Jenkins G (1970) Time series analysis, forecasting and control. Holden-Day, San Francisco
Lee Y-S, Tong L-I (2011) Forecasting time series using a methodology based on autoregressive integrated moving average and genetic programming. Knowl Based Syst 24:66–72
Kihoro JM, Otieno RO, Wafula C (2004) Seasonal time series forecasting: a comparative study of ARIMA and ANN models. Afr J Sci Technol (AJST) Sci Eng Ser 5(2):41–49
Aladag CH, Egrioglu E, Kadilar C (2009) Forecasting nonlinear time series with a hybrid methodology. Appl Math Lett 22:1467–1470
Aydin I, Karakose M, Akin E (2014) An approach for automated fault diagnosis based on a fuzzy decision tree and boundary analysis of a reconstructed phase space. ISA Trans 53:220–229
Veloz A, Salas R, Allende-Cid H, Allende H, Moraga C (2015) Identification of lags in nonlinear autoregressive time series using a flexible fuzzy model. Neural Process Lett 43(3):641–666
Aladag CH, Egrioglu E, Yolcu U (2010) Forecast combination by using artificial neural networks. Neural Process Lett 32:269–276
Xiong T, Bao Y, Hu Z, Chiong R (2015) Forecasting interval time series using a fully complex-valued RBF neural network with DPSO and PSO algorithms. Inf Sci 305:77–92
Kıran MS, Özceylan E, GÜndÜz M, Paksoy T (2012) A novel hybrid approach based on particle swarm optimization and ant colony algorithm to forecast energy demand of Turkey. Energy Convers Manag 53:75–83
Behnamian J, Ghomi SMTF (2010) Development of a PSO–SA hybrid metaheuristic for a new comprehensive regression model to time-series forecasting. Expert Syst Appl 37:974–984
Orfila A, Ballester JL, Oliver R, Alvarez A, Tintoré J (2002) Forecasting the solar cycle with genetic algorithms. Astron Astrophys 386:313–318
Abdullahi M, Ngadi MdA, Abdulhamid SM (2016) Symbiotic organism search optimization based task scheduling in cloud computing environment. Future Gener Comput Syst 56:640–650
Duman S (2017) Symbiotic organisms search algorithm for optimal power flow problem valve-point effect and prohibited zones. Neural Comput Appl 28:3571–3585
Goldberg DE, Holland JH (1988) Genetic algorithms and machine learning. Mach Learn 3:95–99
Kennedy J, Eberhart R (1995) Particle swarm optimization. In: Proceedings IEEE international conference on neural networks, vol 4, pp 1942–1948
Dorigo M, DiCaro G (1999) Ant colony optimization: a new meta-heuristic. In: Proceedings 1999 congress on evolutionary computation, July 6–9, pp 1470–1477
Karaboga D, Basturk B (2007) A powerful and efficient algorithm for numerical function optimization: artificial bee colony (ABC) algorithm. J Glob Optim 39:459–471
Sunspot Index Data Center from the website http://sidc.oma.be/
Mackey MC, Glass L (1977) Oscillation and chaos in physiological control system. Science 197:287–289
Passengers carried in domestic Indian air-lines from the website dgca.nic.in/reports/stat-ind.htm
Homogeneous Indian monthly rainfall data sets from the website www.tropmet.res.in/Data%20Archival-51-Page
Lynx Data (Annual Canadian Lynx Trappings, 1821–1934; listed by columns) from the website galton.uchicago.edu/~pourahm/book/lynxdata.pdf
Bombay stock exchange sensex data collected from the website http://www.bseindia.com/indices/indexarchivedata.aspx
Acknowledgements
The first author would like to thank DST INSPIRE, India, for their help and supports to sustain the work.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
All authors declare that they have no conflicts of interest regarding the publication of this manuscript.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
Five data sets on sunspots, domestic airline passengers, rainfall measurement, number of lynx and BSE sensex closing price have been given in following Tables 6, 7, 8, 9 and 10, respectively.
Functional forms generated from the algorithmic steps for each time series are given in Eqs. (9)–(14).
- 1.
Functional form for the sunspots time series with window size six is given below:
$$\begin{aligned} x[t]&=(4.09\times 10^{-02})/x[t-5]/x[t-3]\nonumber \\&+\,(-1.62)+(-1.34)/x[t-1]*x[t-4]\nonumber \\&+\,(5.126\times 10^{-01})/x[t-1]*x[t-3]+(1.60)/\nonumber \\&x[t-3]+(3.79\times 10^{-01})*x[t-1]/x[t-1]\nonumber \\&+\,(9.92\times 10^{-02})/x[t-3]*x[t-3]\nonumber \\&+\,(1.15)*x[t-1]+(-9.21\times 10^{-01})/\nonumber \\&x[t-2]+(2.11\times 10^{-01})*x[t-3]\nonumber \\&+\,(-9.48\times 10^{-04})*x[t-3]*x[t-4]\nonumber \\&+\,(-4.58\times 10^{-02})*x[t-2]+(2.30\times 10^{-01})/\nonumber \\&x[t-4]+(-2.79\times 10^{-04})*x[t-3]/x[t-2]\nonumber \\&+\,(1.82\times 10^{-03})*x[t-4]*x[t-2]\nonumber \\&+\,(-2.65\times 10^{-04})*x[t-2]/x[t-5]+\nonumber \\&(-9.86\times 10^{-02})*x[t-5]+(2.19\times 10^{-01})*x[t-2]/x[t-3]\nonumber \\&+\,(-1.46\times 10^{-01})/x[t-4]/x[t-2]+(-9.86\nonumber \\&\times 10^{-01})/x[t-5]/x[t-2]+(1.24\times 10^{-01})*x[t-5]/x[t-3]\nonumber \\&+\,(1.04\times 10^{-01})*x[t-4]+(6.04\times 10^{-03})/\nonumber \\&x[t-5]+(-4.56\times 10^{-04})/x[t-4]/x[t-3]\nonumber \\&+\,(-5.73\times 10^{-04})*x[t-4]*x[t-5]\nonumber \\&+\,(-6.21\times 10^{-04})/x[t-4]*x[t-3]\nonumber \\&+\,(6.14\times 10^{-04})*x[t-4]/x[t-2]\nonumber \\&+\,(7.38\times 10^{-04})*x[t-1]*x[t-1]\nonumber \\&+\,(-3.64\times 10^{-03})/x[t-1]+(1.06\times 10^{-04})\nonumber \\&/x[t-1]*x[t-2]+(1.37\times 10^{-04})*x[t-2]*x[t-3]\nonumber \\&+\,(7.02\times 10^{-03})/x[t-1]/x[t-4]+(-1.99\times 10^{-04})/x[t-5]\nonumber \\&/x[t-5]+(1.73\times 10^{-01})/x[t-2]*x[t-2]\nonumber \\&+\,(6.19\times 10^{-01})/x[t-1]*x[t-5]\nonumber \\&+\,(5.05\times 10^{-01})*x[t-4]/x[t-1]+\nonumber \\&(1.97\times 10^{-03})*x[t-4]/x[t-5]\nonumber \\&+\,(5.37\times 10^{-03})/x[t-3]*x[t-4]\nonumber \\&+\,(-1.01\times 10^{-03})/x[t-2]/x[t-5]+\nonumber \\&(-3.06\times 10^{-04})/x[t-3]/x[t-2]\nonumber \\&+\,(-3.47\times 10^{-03})/x[t-5]*x[t-1]\nonumber \\&+\,(-4.89\times 10^{-03})*x[t-2]*x[t-2]+\nonumber \\&(2.80\times 10^{-03})/x[t-5]*x[t-5]\nonumber \\&+\,(-1.43\times 10^{-03})*x[t-5]*x[t-1]\nonumber \\&+\,(1.39\times 10^{-03})*x[t-5]/x[t-1]+\nonumber \\&(-2.90\times 10^{-03})*x[t-1]*x[t-4]\nonumber \\&+\,(1.37\times 10^{-02})/x[t-1]*x[t-1]\nonumber \\&+\,(8.12\times 10^{-04})/x[t-4]/x[t-4]+\nonumber \\&(3.00\times 10^{-04})*x[t-5]*x[t-4]\nonumber \\&+\,(1.38\times 10^{-03})*x[t-3]*x[t-2]\nonumber \\&+\,(7.09\times 10^{-04})*x[t-5]/x[t-5]+\nonumber \\&(1.24\times 10^{-03})*x[t-1]/x[t-3] \end{aligned}$$(9) - 2.
Functional form for the Mackey–Glass time series with window size six is given below:
$$\begin{aligned} x[t]&=(2.02)*x[t-5]+(-9.70\times 10^{-01})/x[t-1]/x[t-4]\nonumber \\&+\,(3.58\times 10^{-01})*x[t-1]+(5.85\times 10^{-01})*x[t-4]+\nonumber \\&(-1.39)/x[t-5]*x[t-3]+(-1.06)*x[t-1]/x[t-4]\nonumber \\&+\,(-2.05\times 10^{-02})*x[t-4]/x[t-1]+(1.51)/x[t-1] \nonumber \\&+\,(-1.22\times 10^{-01})/x[t-2]+(1.45)/x[t-4]+(2.05\times 10^{-02})*x[t-4]\nonumber \\&+\,(1.49\times 10^{-01})*x[t-1]*x[t-3]+(-8.18\times 10^{-01})*\nonumber \\&x[t-3]+(-2.74\times 10^{-01})*x[t-5]/x[t-3]\nonumber \\&+\,(-4.08\times 10^{-01})/x[t-1]/x[t-1]\nonumber \\&+\,(4.61\times 10^{-01})/x[t-4]*x[t-3]+\nonumber \\&(1.67\times 10^{-02})*x[t-5]/x[t-3]+(1.02e+00)/x[t-3]\nonumber \\&+\,(1.07e+00)+(-1.85e-02)/x[t-3]+(-8.20e-01)/\nonumber \\&x[t-3]/x[t-2]+(-1.83\times 10^{-02})\nonumber \\&+\,(6.61\times 10^{-02})*x[t-3]/x[t-1]\nonumber \\&+\,(2.78\times 10^{-01})*x[t-2]*x[t-1]+(-7.98\times \nonumber \\&10^{-01})*x[t-5]*x[t-1]+(1.88)*x[t-3]*x[t-4]\nonumber \\&+\,(-1.02)/x[t-5]+(-1.90\times 10^{-01})*x[t-3]/x[t-4]+(3.91)/\nonumber \\&x[t-4]/x[t-2]+(-7.01\times 10^{-01})/x[t-3]/x[t-3]\nonumber \\&+\,(-2.43)/x[t-2]/x[t-1]\nonumber \\&+\,(-2.79)*x[t-4]/x[t-4]+(-1.73)/\nonumber \\&x[t-1]*x[t-1]+(2.18)*x[t-2]+(-4.70^{e-02})/x[t-4]/x[t-4]\nonumber \\&+\,(-5.30\times 10^{-01})*x[t-1]/x[t-3]+(-8.71\times \nonumber \\&10^{-01})*x[t-4]*x[t-2]+(-2.88\times 10^{-01})/x[t-4]/x[t-5]\nonumber \\&+\,(1.29\times 10^{-01})*x[t-1]/x[t-1]+(-2.31\times 10^{-01})/\nonumber \\&x[t-2]*x[t-4]+(-6.49\times 10^{-01})*x[t-1]*x[t-4]\nonumber \\&+\,(5.29\times 10^{-01})*x[t-1]/x[t-2]+(3.91\times 10^{-01})*x[t-5]/\nonumber \\&x[t-1]+(-8.03\times 10^{-01})/x[t-4]*x[t-2]\nonumber \\&+\,(7.90\times 10^{-01})/x[t-3]/x[t-1]+(6.31\times 10^{-01})/x[t-1]*x[t-5]+\nonumber \\&(-8.15\times 10^{-01})*x[t-4]*x[t-5]\nonumber \\&+\,(3.80\times 10^{-01})/x[t-2]/x[t-5]+(9.79\times 10^{-01})/x[t-2]*x[t-1]\nonumber \\ \end{aligned}$$(10) - 3.
Functional form for the air passengers time series with window size six is given below:
$$\begin{aligned} x[t]&=(-8.16\times 10^{-01})/x[t-4]*x[t-2]\nonumber \\&+\,(-1.30)*x[t-1]/x[t-5]\nonumber \\&+\,(-2.33)/x[t-1]+(1.19\times 10^{-01})*x[t-3]+(9.25\times \nonumber \\&10^{-01})+(-1.66\times 10^{-01})/x[t-2]/x[t-1]+(1.00)*x[t-5]\nonumber \\&+\,(-1.30\times 10^{-02})/x[t-1]+(1.75\times 10^{-03})+\nonumber \\&(1.61\times 10^{-02})+(1.07)/x[t-2]\nonumber \\&+\,(9.11\times 10^{-01})/x[t-2]*x[t-1]\nonumber \\&+\,(4.79\times 10^{-03})*x[t-3]+(-1.03)/x[t-1]/\nonumber \\&x[t-3]+(1.39\times 10^{-03})*x[t-5]+(-3.28\times 10^{-03})\nonumber \\&+\,(-1.32)*x[t-2]/x[t-4]+(-5.35\times 10^{-01})/x[t-4]+\nonumber \\&(-1.07)/x[t-5]*x[t-2]+(1.72)*x[t-2]/x[t-2]\nonumber \\&+\,(-3.13\times 10^{-01})/x[t-4]*x[t-1]\nonumber \\&+\,(-1.66)*x[t-4]+(1.47)/\nonumber \\&x[t-1]*x[t-4]+(2.54\times 10^{-01})/x[t-5]*x[t-5]\nonumber \\&+\,(5.95\times 10^{-01})/x[t-3]/x[t-4]+(-2.32\times 10^{-03})*\nonumber \\&x[t-4]+(1.50)*x[t-2]+(8.36\times 10^{-03})*x[t-2]\nonumber \\&+\,(6.88\times 10^{-01})*x[t-4]/x[t-1]+(-8.11\times 10^{-01})/\nonumber \\&x[t-5]+(-1.74)/x[t-4]/x[t-4]+(9.06\times 10^{-01})/x[t-1]/x[t-1]\nonumber \\&+\,(-2.28\times 10^{-01})*x[t-1]/x[t-1]+\nonumber \\&(3.23\times 10^{-02})*x[t-1] \end{aligned}$$(11) - 4.
Functional form for the rainfall time series window size twelve is given below:
$$\begin{aligned} x[t]&=(-1.43\times 10^{-01})*x[t-8]\nonumber \\&+\,(5.80\times 10^{-03})/x[t-11]\nonumber \\&+\,(-5.29\times 10^{-01})/x[t-4]\nonumber \\&+\,(3.13 \times 10^{-02})*x[t-4]+\nonumber \\&(-2.69\times 10^{-01})/x[t-2]\nonumber \\&+\,(7.30\times 10^{-03})/x[t-9]/x[t-11]\nonumber \\&+\,(4.70\times 10^{-04})*x[t-5]\nonumber \\&+\,(3.57\times 10^{-03})*x[t-1]/x[t-2]+\nonumber \\&(-4.89\times 10^{-03})/x[t-8]\nonumber \\&+\,(-2.26\times 10^{-03})/x[t-4]*x[t-9]\nonumber \\&+\,(1.30\times 10^{-03})*x[t-2]*x[t-9]\nonumber \\&+\,(-1.52\times 10^{-01})/\nonumber \\&x[t-3]+(2.06\times 10^{-04})*x[t-3]\nonumber \\&+\,(-2.94\times 10^{-02})*x[t-7]/x[t-8]\nonumber \\&+\,(2.63\times 10^{-02})+(-1.74\times 10^{-04})*x[t-9]*\nonumber \\&x[t-3]+(-1.22\times 10^{-01})*x[t-11]\nonumber \\&+\,(-1.78\times 10^{-01})/x[t-2]/x[t-8]\nonumber \\&+\,(8.25\times 10^{-02})*x[t-2]+(-4.82\times \nonumber \\&10^{-01})/x[t-11]*x[t-5]\nonumber \\&+\,(-2.60\times 10^{-01})*x[t-1]/x[t-9]\nonumber \\&+\,(3.08\times 10^{-02})*x[t-9]/x[t-1]+(-1.39\times 10^{-01})/\nonumber \\&x[t-7]+(6.88\times 10^{-02})/x[t-10]*x[t-5]\nonumber \\&+\,(-7.09\times 10^{-02})*x[t-6]\nonumber \\&+\,(-8.74\times 10^{-02})/x[t-9]+(7.63\times 10^{-02})*\nonumber \\&x[t-5]/x[t-7]+(1.53\times 10^{-01})*x[t-7]\nonumber \\&+\,(-2.11\times 10^{-02})/x[t-5]\nonumber \\&+\,(5.49\times 10^{-02})/x[t-9]/x[t-3]+\nonumber \\&(-1.07\times 10^{-02})/x[t-6]/x[t-6]\nonumber \\&+\,(-3.48\times 10^{-02})/x[t-5]/x[t-1]\nonumber \\&+\,(-2.18\times 10^{-02})/x[t-2]*x[t-1]+(3.40\times \nonumber \\&10^{-02})*x[t-9]+(-8.42\times 10^{-04})*x[t-2]*x[t-8]\nonumber \\&+\,(-1.10\times 10^{-02})*x[t-1]/x[t-5]+(-1.21\times 10^{-04})*\nonumber \\&x[t-11]/x[t-2]+(1.58\times 10^{-03})/x[t-9]*x[t-4]\nonumber \\&+\,(-9.85\times 10^{-04})/x[t-1]+(-1.04\times 10^{-03})*x[t-1]*x[t-6]+\nonumber \\&(-9.49\times 10^{-04})*x[t-10]/x[t-7]\nonumber \\&+\,(5.05\times 10^{-04})/x[t-1]/x[t-1]\nonumber \\&+\,(-1.25\times 10^{-03})/x[t-7]*x[t-5]+(7.06\times \nonumber \\&10^{-04})*x[t-9]/x[t-6]+(2.00\times 10^{-04})/x[t-6]/x[t-4]\nonumber \\&+\,(7.05\times 10^{-04})*x[t-6]/x[t-2]\nonumber \\&+\,(4.97\times 10^{-04})*x[t-5]*\nonumber \\&x[t-2]+(4.46\times 10^{-04})*x[t-11]*x[t-1]\nonumber \\&+\,(1.41\times 10^{-04})*x[t-9]*x[t-11]\nonumber \\&+\,(1.07\times 10^{-04})/x[t-11]*x[t-7] \end{aligned}$$(12) - 5.
Functional form for the Lynx time series with window size six is given below:
$$\begin{aligned} \begin{aligned} x[t]&=(5.39\times 10^{-01})/x[t-4]/x[t-2]+(1.42)/x[t-3]/x[t-3]\\&+\,(-4.93\times 10^{-01})*x[t-3]+(-2.54)/x[t-2]/x[t-1]+\\&(-4.34)*x[t-4]/x[t-4]+(1.51)/x[t-1]*x[t-1]\\&+\,(-3.07)/x[t-2]+(-2.11\times 10^{-01})*x[t-2]\\&+\,(5.46\times 10^{-01})/\\&x[t-1]+(1.97\times 10^{-01})*x[t-5]\\&+\,(2.02)+(3.76\times 10^{-01})/x[t-5]*x[t-3]\\&+\,(3.19\times 10^{-01})*x[t-4]+(3.51)/\\&x[t-4]+(1.33)*x[t-3]/x[t-4]\\&+\,(1.83)/x[t-3]+(-3.24)/x[t-1]/x[t-4]\\&+\,(1.15)*x[t-1]+(-3.60)*x[t-1]/ \\&x[t-4]+(-9.90\times 10^{-01})*x[t-4]/x[t-1]\\&+\,(3.08\times 10^{-07})*x[t-4]/x[t-1]\\&+\,(-1.56\times 10^{-04})/x[t-4]/x[t-4]+ \\&(-1.46\times 10^{-04})/x[t-5]\\&+\,(-1.29\times 10^{-04})/x[t-4]/x[t-5]\\&+\,(-1.52\times 10^{-04})*x[t-1]*x[t-4]\\&+\,(-1.35\times 10^{-03})*x[t-1]/x[t-2]\\&+\,(1.94\times 10^{-04})/x[t-4]*x[t-2]\\&+\,(-8.19\times 10^{-04})/x[t-3]/x[t-1] \end{aligned} \end{aligned}$$(13) - 6.
Functional form for the BSE sensex closing price time series with window size six is given below:
$$\begin{aligned} \begin{aligned} x[t]&=(4.60\times 10^{-01})/x[t-2]+(7.01\times 10^{-02})*x[t-1]\\&+\,(-3.45\times 10^{-01})/x[t-1]*x[t-1]+(4.95\times \\&10^{-01})/x[t-3]+(1.45\times 10^{-01})/x[t-1]\\&+\,(6.39\times 10^{-01})/x[t-5]*x[t-3]+(2.42\times 10^{-01})+\\&(2.23\times 10^{-01})*x[t-4]+(-3.19\times 10^{-01})*x[t-3]\\&+\,(-9.19\times 10^{-02})/x[t-4]+(-1.37\times 10^{-01})*\\&x[t-3]/x[t-4]+(6.85\times 10^{-01})*x[t-5]\\&+\,(1.92\times 10^{-01})/x[t-1]/x[t-4]\\&+\,(3.57\times 10^{-01})*x[t-1]/x[t-4]+\\&(2.61\times 10^{-02})*x[t-4]/x[t-1]\\&+\,(-3.23\times 10^{-01})*x[t-4]/x[t-4]\\&+\,(3.71\times 10^{-01})*x[t-2]+\\&(5.59\times 10^{-01})/x[t-4]/x[t-2]\\&+\,(-3.18\times 10^{-01})/x[t-3]/x[t-3]\\&+\,(-4.12\times 10^{-01})/x[t-2]/x[t-1]+\\&(1.30\times 10^{-04})*x[t-1]/x[t-2] \end{aligned} \end{aligned}$$(14)
We have neglected the terms with coefficient value less than \(10^{-04}\) for time series sunspots, air passengers, rainfall, lynx data, BSE sensex and coefficient value less than \(10^{-02}\) for Mackey–Glass in all the above expressions in our writing but still our computations remain unaffected.
Rights and permissions
About this article
Cite this article
Pal, S.S., Samui, S. & Kar, S. A new technique for time series forecasting by using symbiotic organisms search. Neural Comput & Applic 32, 2365–2381 (2020). https://doi.org/10.1007/s00521-019-04134-8
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00521-019-04134-8