Abstract
This paper brings together two important developments in forecasting literature; the artificial neural networks and factor models. The paper introduces the factor augmented artificial neural network (FAANN) hybrid model in order to produce a more accurate forecasting. Theoretical and empirical findings have indicated that integration of various models can be an effective way of improving on their predictive performance, especially when the models in the ensemble are quite different. The proposed model is used to forecast three time series variables using large South African monthly panel, namely, deposit rate, gold mining share prices and Long term interest rate, using monthly data over the in-sample period (training set) 1992:1–2006:12. The variables are used to compute out-of-sample (testing set) results for 3, 6 and 12 month-ahead forecasts for the period of 2007:1–2011:12. The out-of-sample root mean square error findings show that the FAANN model yields substantial improvements over the autoregressive AR benchmark model and standard dynamic factor model (DFM). The Diebold–Mariano test results also further confirm the superiority of the FAANN model forecast performance over the AR benchmark model and the DFM model forecasts. The superiority of the FAANN model is due to the ANN flexibility to account for potentially complex nonlinear relationships that are not easily captured by linear models.
Similar content being viewed by others
Notes
The approach works as follows; Suppose a factor model is represented as \(X_{it} =\lambda ^{\prime }_i F_t +e_{it}\) where \(X_{it}\) is the observed datum for the \(i\mathrm{th}\) series at time \(t\,(i=1,\ldots , N; t=1,\ldots ,T); F_t \) is a vector \((r\times 1)\) of common factors; \(\lambda _i\) is a vector \(\left( {r\times 1} \right) \) of factor loadings; and \(e_{it}\) is the idiosyncratic component of \(X_{it}\). The right hand side variables are not observed. The method of principal components minimizes \(V\left( r \right) =\textit{min}_{{\Lambda }, F} \left( {NT} \right) ^{-1}\mathop {\sum }_{i=1}^N \mathop {\sum }_{t=1}^T \left( {X_{it} -\lambda ^{\prime }_i F_t} \right) ^{2}\) where \(\varLambda =\left( {\lambda _1, \ldots ,\lambda _N} \right) \). Concentrating out \(\Lambda \) and using the normalization that \(F^{\prime }F/T=I_r\), where \(I_r\) is \(r\times r\) identity matrix, the problem is identical to maximizing \(\textit{tr}\left( {{F}^{\prime }}(XX^{\prime })F\right) \). The estimated factor matrix, denoted by \(\tilde{F}\), is \(\sqrt{T}\) times the eigenvectors corresponding to the r largest eigenvalues of the \(T\times T\) matrix \(X^{\prime }X\), and \(\tilde{\varLambda }^{\prime }=(\tilde{F}^{\prime }\tilde{F})^{-1}\tilde{F} X=\tilde{F} X/T\) is the corresponding loading matrix.
In this paper we choose iterated forecast instead of direct forecast. Marcellino et al. [27] found that iterated forecast using AIC lag length selection performed better than direct forecasts, especially when forecast horizon increases. They argued that iterated forecast models with lag length selected based on information criterion are good estimates for the best linear predictor.
The data sources are the South Africa Reserve Bank, ABSA Bank, Stats South Africa, National Association of Automobile Manufacturers of South Africa (NAAMSA), South African Revenue Service (SARS), Quantec and World Bank.
The RMSE statistic can be defined as \(\sqrt{\frac{1}{N}{\sum }\left( {Y_{t+n} -{}_{t} \hat{Y}_{t+n}}\right) ^{2}}\), where \(Y_{t+n}\) denotes the actual value of a specific variable in period \(t+n\) and \({}_t\hat{Y}_{t+n}\) is the forecast made in period t for \(t+n.\)
References
Aruoba S, Diebold F, Scotti C (2009) Real-time measurement of business conditions. J Bus Econ Stat 27:417–427
Bai J (2003) Inferential theory for factor models of large dimensions. Econometrica 71(1):135–171
Bai J, Ng S (2002) Determining the number of factors in approximate factor models. Econometrica 70(1):191–221
Bai J, Ng S (2007) Determining the number of primitive shocks. J Bus Econ Stat 25(1):52–60
Bai J, Ng S (2010) Instrumental variable estimation in a data-rich environment. Econom Theory 26(6):1577–1606
Banerjee A, Marcellino M, Masten I (2008) Forecasting macroeconomic variables using diffusion indexes in short samples with structural change. In: Rapach D, Wohar M (eds) Forecasting in the presence of structural breaks and model uncertainty. Emerald Group, Bingley
Banerjee A, Marcellino M (2008) Factor augmented error correction models. CEPR Discussion Paper, 6707
Baxt WG (1992) Improving the accuracy of an artificial neural network using multiple differently trained networks. Neural Comput 4:772–780
Bernanke B, Boivin J, Eliasz P (2005) Measuring the effects of monetary policy: a factor-augmented vector autoregressive (FAVAR) approach. Q J Econ 120:387–422
Boivin J, Giannoni M (2006) DSGE models in a data-rich environment. Technical Report, Columbia Business School, Columbia University
Chamberlain G, Rothschild M (1983) Arbitrage factor structure and mean-variance analysis in large markets. Econometrica 51:1305–1324
Chen A, Leung MT, Hazem D (2003) Application of neural networks to an emerging financial market: forecasting and trading the Taiwan Stock Index. Comput Oper Res 30:901–923
Diebold FX, Mariano RS (1995) Comparing predictive accuracy. J Bus Econ Stat 13:253–263
Dufour JM, Pelletier D (2013) Practical methods for modelling weak VARMA processes: identification, estimation and specification with a macroeconomic application. Discussion Paper
Favero C, Marcellino M, Neglia F (2005) Principal components at work: the empirical analysis of monetary policy with large datasets. J Appl Econom 20:603–620
Forni M, Hallin M, Lippi M, Reichlin L (2000) The generalized factor model: identification and estimation. Rev Econ Stat 82:540–554
Forni M, Hallin M, Lippi M, Reichlin L (2004) The generalized factor model: consistency and rates. J Econom 119:231–255
Forni M, Hallin M, Lippi M, Reichlin L (2005) The generalized dynamic factor model, one sided estimation and forecasting. J Am Stat Assoc 100(471):830–840
Geweke J (1977) The dynamic factor analysis of economic time series. In: Aigner DJ, Goldberger AS (eds) Latent variables in socio-economic models. North Holland, Amsterdam, pp 365–383
Giannone D, Reichlin L, Small D (2008) Nowcasting: the real-time informational content of macroeconomic data. J Monet Econ 55:665–676
Greg T, Hu S (1999) Forecasting GDP growth using artificial neural networks. Working Paper 3, Bank of Canada
Hallin M, Liska R (2007) Determining the number of factors in the general dynamic factor model. J Am Stat Assoc 102:603–617
Jorge N, Wright Stephen J (2006) Numerical optimization, 2nd edn. Springer, Berlin
Kapetanios G, Marcellino M (2010) Factor-GMM estimation with large sets of possibly weak instruments. Comput Stat Data Anal 54:2655–2675
Khashei M, Bijari M (2010) An artificial neural network (p, d, q) model for time series forecasting. Expert Syst Appl 37:479–489
Kihoro JM, Otieno RO, Wafula C (2004) Seasonal time series forecasting: a comparative study of ARIMA and ANN models. Afr J Sci Technol 5(2):41–49
Marcellino M, Stock JH, Watson MW (2006) A comparison of direct and iterated multistep AR methods for forecasting macroeconomic series. J Econom 135:499–526
Ng S, Stevanovic D (2012) Factor augmented autoregressive distributed lag models. Mimeo, Columbia University, New York
Onatski A (2009) Testing hypotheses about the number of factors in large factor models. Econometrica 77:1447–1479
Philip AA, Taofiki AA, Bidemi AA (2011) Artificial neural network model for forecasting foreign exchange rate. World Comput Sci Inf Technol J 1(3):110–118
Sargent TJ, Sims CA (1977) Business cycle modeling without pretending to have too much a priori economic theory. In: Sims C (ed) New methods in business research. Federal Reserve Bank of Minneapolis, Minneapolis
Stock JH, Watson MW (2005) Implications of dynamic factor models for VAR analysis. Manuscript. Princeton University, Princeton
Stock JH, Watson MW (2002a) Forecasting using principal components from a large number of predictors. J Am Stat Assoc 97:147–162
Stock JH, Watson MW (2002b) Macroeconomic forecasting using diffusion indexes. J Bus Econ Stat 20:147–162
Tseng FM, Yu HC, Tzeng GH (2002) Combining neural network model with seasonal time series ARIMA model. Technol Forecast Soc Change 69:71–87
Yu L, Wang S, Lai K (2005) A novel nonlinear ensemble forecasting model incorporating GLAR and ANN for foreign exchange rates. Comput Oper Res 32:2523–2541
Zhang G, Patuwo BE, Hu MY (1998) Forecasting with artificial neural networks: the state of the art. Int J Forecast 14:35–62
Zhang GP (2003) Time series forecasting using a hybrid ARIMA and neural network model. Neurocomputing 50:159–175
Zhang GP (2007) A neural network ensemble method with jittered training data for time series forecasting. Inf Sci 177:5329–5346
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Babikir, A., Mwambi, H. Factor Augmented Artificial Neural Network Model. Neural Process Lett 45, 507–521 (2017). https://doi.org/10.1007/s11063-016-9538-6
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11063-016-9538-6