Study of Discrete Choice Models and Adaptive Neuro-Fuzzy Inference System in the Prediction of Economic Crisis Periods in USA

Abstract

Abstract:In this study two approaches are applied for the prediction of the economic recession or expansion periods in USA. The first approach includes Logit and Probit models and the second is an Adaptive Neuro-Fuzzy Inference System (ANFIS) with Gaussian and Generalized Bell membership functions. The in-sample period 1950-2006 is examined and the forecasting performance of the two approaches is evaluated during the out-of sample period 2007-2010. The estimation results show that the ANFIS model outperforms the Logit and Probit model. This indicates that neuro-fuzzy model provides a better and more reliable signal on whether or not a financial crisis will take place.

An Adaptive Neuro-Fuzzy Inference System for Forecasting Australia’s Domestic Low Cost Carrier Passenger Demand

“…A further advantage of the ANFIS is the fact that it can be trained without the requirement for the expert knowledge normally required for the standard fuzzy logic design, and both numerical and linguistic knowledge can be combined into a fuzzy rule base by utilising fuzzy methods (Giovanis 2012). Other important advantages of the ANFIS include its nonlinear ability, its capacity for rapid learning, and its adaptation capability.…”

“…Furthermore, the strength of the ANFIS is that it uses the artificial neural network's ability to classify data and identify patterns. Moreover, the ANFIS develops a fuzzy expert system that is more transparent to the user and which is also less likely to produce memorization errors than an ANN (Giovanis 2012).…”

“…The fifth ANFIS layer, whose node is labelled "∑", is the output layer, in which a single node calculates the overall output as a summation of all incoming signals (Ch, Mathur 2010;Giovanis 2012;Xiao et al 2014). Hence, the overall output of the model can be written as (Fang 2012;yetilmezsoy et al 2011):…”

“…Since the ANFIS is an adaptive network which permits the use of ANN topology as well as fuzzy logic, it includes the characteristics of both and also eliminates some disadvantages of these methods when used individually (yetilmezsoy et al 2011). Accordingly, the ANFIS is capable of handling complex and nonlinear problems (Giovanis 2012). Even if the targets are not given, the ANFIS may reach the optimum result very quickly (yetilmezsoy et al 2011).…”

“…Consequently, artificial intelligence-based modelling techniques have become more popular in diverse disciplines over the past decade (Kar et al 2014) because of their robustness, high predictive capabilities and flexible behaviours to handle the multi-objective criteria in a straightforward manner (yetilmezsoy et al 2011). Jang (1993) and Jang et al (1997) introduced the adaptive network-based fuzzy inference system (AN-FIS), which is a system using a hybrid learning rule to optimize the fuzzy system parameters of a first order Sugeno system (Giovanis 2012). This approach has been applied to a growing range of disciplines, including transport mode choice (Andrade et al 2007), economics (Fang 2012;Giovanis 2012), electricity demand forecasting (Zahedi et al 2013), financial markets forecasting (Bagheri et al 2014;Kablan 2009), gold price forecasting (Makridou et al 2013), oil consumption forecasting (Senvar et al 2013), stock market forecasting (Atsalakis, Valavanis 2009; Chen et al 2013;Cheng et al 2013;Svalina et al 2013;Wei 2013), tourism demand forecasting (Atsalakis et al 2014;Chen et al 2010;Hadavandi et al 2011), and ordering policy in supply chains (Latif et al 2014).…”