Comparing the Effectiveness of Predict Model for Price Return of World Gold by Employing Bootstrap and Artificial Neural Network

  • ภูริพงศ์ อัครานุชาต
  • สมพร ปั่นโภชา
  • บำรุง พ่วงเกิด
Keywords: Forecast, Bootstrap, Artificial Neural Network


The objective of this study is to compare the forecasting performance of the world gold price return by the bootstrap method and the artificial neural network. Information used in the study is the world gold price closed daily and technical information. Throughout the period from January 1, 2007 to December 31, 2018, a total of 3,120 days. The study has divided the data into two groups: 2,860 days for modeling and 260 days for the rest created a model to generate predictions for gold price returns and then compare the efficiency of the model.
In the study of forecasting of bootstrap on theoretical calculations which uses a sample that is consistent in the past representative samples in helping to reliably calculations the statistical features of the population. It is as a non-parameter model. The bootstrap variables from error vector and simulation for 10,000 iterations have been performed in order to find the return distribution of gold price returns. Results of the forecasting from the bootstrap model with the root mean square error (RMSE) of the best of 5 models, the values are 43.8874, 49.4084, 49.8692, 47.0801, 51.2655 respectively and the minimum absolute percentage errors (MAPE) are 2.93, 3.20, 3.04, 3.34, 3.37, respectively.
Artificial neural network method consists of input layer, hidden layer and output layer employing learning back propagation method for high efficiency prediction. It is a non-parameter model by using the learning principles to create the relationship of the weight values of the structure within the network. The model is flexible and effective in creating the relationship of the data. The method with 2 hidden layers is enough to solve various problems and number of neural in each hidden layer is moreover affecting the efficiency of calculations. The results from the model predicts the root mean square errors (RMSE) of the top 5 are 14.6785, 15.6151, 16.3905, 19.0690, 23.5769, respectively. Percentages of the minimum absolute errors (Mean absolute percentage errors: MAPE) are at least 0.93, 0.98, 1.13, 1.37, 1.52, respectively.
On the test results in comparing the forecast of the world gold price return between the bootstrap model and artificial neural network models, the best model of each method is comparing according to the criteria by using RMSE, MAPE and MAE values. The criteria RMSE: 43.88, MAPE: 2.93 and MAE: 37.32 is for the bootstrap model and RMSE: 15.61, MAPE: 0.93 and MAE:12.07 for the artificial neural network model. In this study The artificial neural network model has better performance with RMSE, MAE and MAPE values better than the bootstrap model according to the performance measurement criteria of the model for predicting world gold price returns.


Most read articles by the same author(s)

1 2 3 4 5 6 > >>