Gold price forecasting research based on an improved online extreme learning machine algorithm
- PDF / 1,814,061 Bytes
- 11 Pages / 595.276 x 790.866 pts Page_size
- 1 Downloads / 199 Views
ORIGINAL RESEARCH
Gold price forecasting research based on an improved online extreme learning machine algorithm Futian Weng1 · Yinhao Chen1 · Zheng Wang1 · Muzhou Hou1 · Jianshu Luo2 · Zhongchu Tian3 Received: 11 May 2019 / Accepted: 3 January 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract Accurate gold price prediction is highly essential for economic and currency markets. Thus, the intelligence prediction models need to be applied to price prediction. On the basis of long-term collected daily gold, the study proposes a novel genetic algorithm regularization online extreme learning machine (GA-ROSELM), to predict gold price data which collected from public websites. Akaike Information Criterion (AIC) is introduced to build the eight input combinations of variables based on the silver price of the previous day (Silver_D1), Standard & Poor. The 500 indexes (S&P_D1), the crude oil price (Crude_D1), and the gold price of the previous 3 days (Gold_D1, Gold_D2, Gold_D3). Eight optimal variable models are established, and the final input variables are determined according to the minimum AIC value. The proposed model (GAROSELM) solve the problem that OS-ELM model which is easy to generate singular matrices, meanwhile, experiments demonstrate this model performs better than ARIMA, SVM, BP, ELM and OS-ELM in the gold price prediction experiment. On the test set, the root means square error of this model (GA-ROSELM) prediction compared with five other models which decreased by 13.1%, 22.4%, 53.87%, 57.84% and 37.72% respectively. In summary, the results clearly confirm the effectiveness of the GA-ROSELM model. Keywords Genetic algorithm · AIC criterion · Online learning machine · Gold price forecast
1 Introduction As a vital commodity in the economic and currency markets, metal has received much attention in the academic field, especially the gold (Guihao et al. 2010; Baur et al. 2016). In the recent years, to overcome the impact of the economic crisis, many countries and scholars put more efforts to predict the price of gold which makes gold have much attention in the academic field (Zhang and Liao 2014; Bialkowski et al. 2015; Zhong et al. 2019). There are many statistical methods for price forecasting such as AR (Stock and Watson 1988), ARMA (Xu 2017), GARCH (Kristjanpoller and Minutolo 2015), SVM * Muzhou Hou [email protected] 1
School of Mathematics and Statistics, Central South University, Hu Nan Changsha 410083, China
2
College of Science, National University of Defense Technology, Hu Nan Changsha 410073, China
3
School of Civil Engineering, Changsha University of Science and Technology, Changsha 410114, China
(Mustaffa et al. 2015), BP neural network (Yu et al. 2018), and so on. Ye et al. (2019) proposed a new information fusion method of forecasting an made an application to oil price forecast. Kim and Moon (2019) use BiLSTM model for forecasting trading area based on multivariate time series data. Fusion model of wavelet transform and adaptive neuro fuzzy inference
Data Loading...
