Forecasting Gold Price Changes by Using Adaptive Network Fuzzy Inference System

Abstract

Developing a preci se and accurate model of gold price is critical to assets management because of its unique features. In this paper, adaptive neuro-fuzzy inference system (ANFIS) and artifi cial neural network (ANN) model have been used for modeling the gold price, and compared with the traditional statistical model of ARIMA (autoregressive integrated moving average). The three performance measures, the coeffi cient of determination (R 2 ), root mean squared error (RMSE), mean absolute error (MAE), are utilized to evaluate the performances of different models developed. The results show that the ANFIS model outperforms other models (i.e. ANN and ARIMA model), in terms of different performance criteria during the training and validation phases. Sensitivity analysis showed that the gold price changes are highly dependent upon the values of silver price and oil price.

Sensitivity Analysis of Economic Variables using Neuro-Fuzzy Approach

“…Gold price model is highly dependent upon the silver price. Therefore, the proposed model veri es the results reported by Yazdani-Chamzini et al (2012) [9] about gold and silver dependency.…”

“…Some mathematically based SA methods have been proposed in this regard. For example, Yazdani-Chamzini et al [9] applied the Cosine Amplitude Method (CAM) to nd the most sensitive parameters. Although CAM and other mathematically based SA methods applied in the referenced studies [10{14] can be used for economic time series [15], they cannot consider the behavioral similarities of parameters.…”

“…Therefore, they cannot be a proper method in the case of economic variables. On the other hand, learning-based SA methods [9,16] can learn the interactions among economic parameters and, consequently, can better assess the in uence of parameters on the output of an economic model. For example, Valdivia and Arturo [16] applied an Arti cial Neural Network (ANN) for SA.…”

A Model for Shovel Capital Cost Estimation, Using a Hybrid Model of Multivariate Regression and Neural Networks

“…The mean absolute percentage error contains positive numbers without having an upper bound [55,56]. The normalized mean square error presents non-negative numbers, which can be any numbers between zero and one, while the value of 0 means that the model is ideal and the value of one denotes that the constructed model is non-ideal.…”

“…Based on the concepts of the CAM method, the sensitivity for each independent component can be determined through establishing the degree of the relationship (r ij ) between the shovel capital cost and the considered independent component [56]. The larger the value of CAM, the higher its impact on the capital cost.…”