Barış Aşıkgil

Nonlinear time series forecasting with Bayesian neural networks

The Bayesian learning provides a natural way to model the nonlinear structure as the artificial neural networks due to their capability to cope with the model complexity. In this paper, an evolutionary Monte Carlo (MC) algorithm is proposed to train the Bayesian neural networks (BNNs) for the time series forecasting. This approach called as Genetic MC is based on Gaussian approximation with recursive hyperparameter. Genetic MC integrates MC simulations with the genetic algorithms and the fuzzy membership functions. In the implementations, Genetic MC is compared with the traditional neural networks and time series techniques in terms of their forecasting performances over the weekly sales of a Finance Magazine.

A new hybrid method for time series forecasting: AR–ANFIS

In this study, a new hybrid forecasting method is proposed. The proposed method is called autoregressive adaptive network fuzzy inference system (AR–ANFIS). AR–ANFIS can be shown in a network structure. The architecture of the network has two parts. The first part is an ANFIS structure and the second part is a linear AR model structure. In the literature, AR models and ANFIS are widely used in time series forecasting. Linear AR models are used according to model-based strategy. A nonlinear model is employed by using ANFIS. Moreover, ANFIS is a kind of data-based modeling system like artificial neural network. In this study, a linear and nonlinear forecasting model is proposed by creating a hybrid method of AR and ANFIS. The new method has advantages of data-based and model-based approaches. AR–ANFIS is trained by using particle swarm optimization, and fuzzification is done by using fuzzy C-Means method. AR–ANFIS method is examined on some real-life time series data, and it is compared with the other time series forecasting methods. As a consequence of applications, it is shown that the proposed method can produce accurate forecasts.

Polynomial tapered two-stage least squares method in nonlinear regression

Nonlinear models play an important role in various scientific disciplines and engineering. The parameter estimation of these models should be efficient to make better decisions. Ordinary least squares (OLS) method is used for estimating the parameters of nonlinear regression models when all regression assumptions are satisfied. If there is a problem with these assumptions, OLS fails to give efficient results. This paper examines the efficiency of parameter estimation under the problem of autocorrelated errors. Some methods have been proposed in order to overcome the problem and obtain efficient parameter estimates especially for autoregressive (AR) processes. One of the most commonly used method is two-stage least squares (2SLS). This method is based on generalized least squares. In this paper, a novel approach is proposed for 2SLS method by evaluating a polynomial tapering procedure on autocorrelated errors. This new method is called tapered two-stage least squares (T2SLS). The finite sample properties and improvements of T2SLS are explored by means of some real life examples and a Monte Carlo simulation study. Both numerical and experimental results reveal that T2SLS can give more efficient parameter estimates especially in small samples under the autocorrelation problem when compared to OLS and 2SLS.

A Novel Approach for Estimating Seemingly Unrelated Regressions with High-Order Autoregressive Disturbances

A seemingly unrelated regression (SUR) model is defined by a system of linear regression equations in which the disturbances are contemporaneously correlated across equations. However, the disturbances can also be serially correlated in each equation of the system. In these cases, estimating SUR becomes more complicated. Some methods have been considered estimating SUR with low-order autoregressive (AR) disturbances. In this article, SUR with high-order AR disturbances are considered and a tapering approach is examined under this situation. Two modified methods for estimating SUR are obtained by using this approach. A comprehensive Monte Carlo simulation study is performed in order to compare small-sample efficiencies of the modified methods with the others given in the literature.