Dissertation > Excellent graduate degree dissertation topics show

The Combination Forecasting of Bayesian Model Averaging Base on the Maximum Likelihood and Using It in Coal Demand Forecasting

Author: ChenBao
Tutor: DongJingRong
School: Chongqing Normal University
Course: System theory
Keywords: Bayesian model averaging Combination Forecast Posterior probability Maximum Likelihood Coal demand
CLC: F426.21
Type: Master's thesis
Year: 2011
Downloads: 255
Quote: 0
Read: Download Dissertation

Abstract


Coal is China's most important energy sources, coal plays a vital role in the national life, the shortage of coal supply will affect the stability and development of the national economy, coal excess will spread to the healthy development of the coal industry, coal production and meet the requirements for the ultimate pursuit is the development of the market economy. Forecast coal demand is conducive to accelerate the scientific development of China's coal industry, but also conducive to the adjustment of China's energy structure, of great significance to the sustainable development of China's social and economic, scientific and effective. Coal demand forecast is integrated departure from existing coal, economic, social systems, analysis of historical data, to explore the demand for coal and its impact on the law and the relationship between the factors to predict the future demand for coal. Coal demand forecast can be broadly divided into two categories: one is the use of a single model to predict, the second is the use of a combination of model predictions. Due to the complexity of the system of the demand for coal and non-linear characteristics of a single model can not predict it, it is necessary to use a combination forecasting method. Combination Forecasting comprehensive utilization of valid information provided by the individual models, thus improving the prediction accuracy of the method key is to determine the weight of each individual prediction model; combined weights can be divided into constant weight and time-varying weights. Constant weight method for determining the more mature, but the same weight combination forecasting method is difficult to adapt to the reality forecast in the midst of characteristics, can not be reasonably reactive single forecasting model \Time-varying weights change over time and change, can effectively reflect the changes of individual prediction model, time-varying weights research started relatively late, is more difficult to determine. Bayesian method can effectively determine the variable weight, it can be clearly said that the information update process, strong self-adaptability and dynamic adjustment, able to adapt to time-varying weights change over time. The Bayesian combination forecasting method can not only take full advantage of each model is the actual information, and can effectively combine subjective information with the model or data information. This dynamic profile model at the right weight is important. In many Bayesian combination forecasting model is the most typical combination forecasting method of Bayesian model, the posterior probability of each alternative model as a weight to the weighted average of the individual predictive value of all alternate model, resulting combination forecasting estimated value; consider all possible single model, and after posterior probability that the right to update the weight as standard to judge the merits of the model in order to effectively deal with the uncertainty of the model. In this paper, the Bayesian model combination forecasting method based on marginal likelihood of further expansion of the use of alternative marginal maximum likelihood likelihood calculation of the posterior probability of the individual model, to overcome the problem of marginal likelihood over-reliance on prior information. The same time, given the sample data are divided into feature samples and monitoring samples, the Bayesian model combination forecasting method based on maximum likelihood timely dynamic updates right combination of weight great significance. Finally, the use of the combination forecasting short-term demand for coal. The results show that the outstanding single model is given a higher weight, maximum likelihood to determine the combination of weight and Model Selection of the asymptotic property, and the relatively high prediction accuracy.

Related Dissertations

  1. Based on statistics of the lognormal distribution heteroscedasticity model inferred,O212.1
  2. Research and Design of Electronic Equalization Based on Most Likelihood Sequence Estimation,TN911.5
  3. Study on Forecasting Method of Regional Freight Transportation Demand,F224;U492.313
  4. The Application of Mathematical Model in the Prediction of College Enrollment,G647.32
  5. Support Vector Machine in Guiyang Freezing Weather forecast,P456.9
  6. Bayesian Model Averaging Base on the Predictive Likelihood and Using It in Power Load Forecasting,F426.61
  7. The Research on the Supply-Demand Situation and Developing Trend of Coal Industry in China,F426.21
  8. A Random Weighted Linear Estimator of the ARCH Parameters,F830
  9. Adaptive Evolution Analysis of Rotavirus VP7&NSP4 Protein,R373
  10. Blind Sources Separation Algorithm Study and Application by Maximum Likelihood Estimation,TN911.7
  11. The Modeling and Predictive Analysis of Chinese Coal Demand,F426.21
  12. The Study of the Volatility Correlation of Chinese Stock Markets Based on Tree Structured Dcc_multivariate GARCH Model,F832.51
  13. For acoustic array network node localization algorithm research,TN929.5
  14. Analysis on Effects of Power Consumption Orderly of Users on Fired-coal Need,F426.61
  15. MCMC-based Bayesian model averaging Combination Forecasting Method and Its Application in China's energy consumption prediction research,F206
  16. Research on the Technology of the Doppler Shift Estimation in the High Dynamic Environment,TN929.5
  17. Method and Implementation of Height-finding Processing in VHF Radar,TN958
  18. The Study of Dynamic Term Structure Models of China,F822.0
  19. Research on Carrier Synchronization in OFDM Systems,TN919.3
  20. Extended Switching Regression Models with Time-varying Probabilities for Combining Forecasts in Stock Market Volatility,F832.51

CLC: > Economic > Industrial economy > China Industrial Economy > Industrial sector economy
© 2012 www.DissertationTopic.Net  Mobile