A Systematic Optimization for Grey Forecasting Model

• Main Authors: Phan Van Thanh, 潘文成
• Other Authors: Wang Chia-Nan
• Format: Others
• Language: en_US
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/02247515081171166340

博士 === 國立高雄應用科技大學 === 機械與精密工程研究所 === 104 === Grey forecasting is a dynamic model and have been validated, widely used in various fields. In recent years, many scholars have been proposed new procedures and have been achieved promising results aim to improve precision accuracy of GM (1, 1) model.However, the prediction accuracy of GM (1, 1) existing may not be always satisfactory in different scenarios. For example, the data have significant fluctuation, and highly nonlinear, or with lots of noise. Therefore, it is necessary to put forward a systematic approach in order to improve the prediction performance as well as to overcome the restriction existing in the grey forecasting model. In order to deal with these issues, this dissertation combined the mathematical algorithm of the grey forecasting model with the excellent ideas of some previous highlight studies (Zeng et al. 2011; Tan, 2000; Ou, 2012; Truong and Ahn, 2012; Wu et al. 2013, 2015; Huang and Lee, 2011; Shen and Qin, 2014; Zhou et al. 2006; Kayacan et al. 2010; Wang and Hsu, 2008; Lin et al. 2009, etc.) to establish the new systematic optimization approach for improving the prediction performance. The new systematic approach is proposed to improve its performance by the following five aspects. The first one is used two smartly additive factors (c1 and c2) to convert any raw data into a grey sequence which satisfies both the raw data checking condition and quasi-smooth condition to perform the grey estimation. The second one is used moving average operation method on the original sequence to smooth the raw sequence data. The third one is a modification in calculating the background value to eliminate the error term. The fourth one is establishing an optimization equation to find the optimal parameters by Genetic algorithm (GA) in the grey forecasting model. The final, we adopt residual error modification methods to reduce the periodic residual error.To verify the effectiveness of the proposed model, both fluctuation data of the numerical example in papers (Ou, 2012; Wang and Hsu, 2008) and some practical applications are used. These all simulation results demonstrated that the proposed approach could offer a more precise forecast than several different kinds of grey forecasting models. For future direction, this proposed model can be applied to forecast the performance with high fluctuation data in the different industries.