Font Size: a A A

The Research Of Uncertain Multi-attribute Decision-making Problems Based On Entropy And Error-eliminating Theory

Posted on:2016-07-10Degree:DoctorType:Dissertation
Country:ChinaCandidate:D R LuoFull Text:PDF
GTID:1109330461457023Subject:Management Science and Engineering
Abstract/Summary:PDF Full Text Request
At the 21st Century, the advance of the society, the change of the environment and the expansion of the information make the scale of decision become much larger, the uncertainty of decision increased and decision-making more difficultly than ever. Some methods exist limitation when making decision under new conditions. Thus decision-making methods should be improved with the times to adapt the changes. Meanwhile, with the progress of science and technology, new theories, new technologies and new methods appear and become efficient tools for decision-making under uncertain conditions. Considering the above, some problems and applications of decision-making and forecasting under uncertain conditions using entropy, error-eliminating theory and SVM are researched in this thesis.(1) Scientific decision can be made without accurately understanding the trends of development. The complex exterior environment which changes makes nonlinear characteristic of data obvious. Furthermore the accuracy of forecasting can’t be ensured when the time period is long. RAR model, fitting residual to improve the accuracy, has good ability to extract the linear information included in time series data and can be interpreted easily. Due to its own characteristics it has weak ability to process the nonlinear information. SVR has strong nonlinear information processing and generalization ability. But it has the deficiency of parameters setting difficultly. Thus revising RAR residual by SVR, a forecasting model is proposed in the thesis. The model, taking advantages of RAR and SVR, overcoming their shortcomings, can predict different period data series having linear and nonlinear characters. It can support decision-making under uncertain conditions.(2) Intuitionistic fuzzy set extends normal fuzzy set and express the assessing information exactly. Yet the uncertainty is increasing when using IFS. To improve the scientific of this type of decision problem, the uncertain measure of IFS information is studied in the thesis. According to the deficiency of existing methods for computing IFS entropy, a new improved method is proposed. The objective weight is determined using IFS entropy. Revising the bias of decision-maker, integrating the IFS information with different operators, a comprehensive model is built and applied to human resource selection decision in this thesis.(3) Current decision-making methods, pursuing the maximum benefit blindingly, will lead to ignore avoiding the loss when decision is made wrongly. Hybrid multi-attributes decision-making based on error-eliminating theory will be researched in the thesis. Analyzing the effect of different attributes in decision-making, studying the relation between error functions and attributes with different functions (crucial, important, redundant and etc.) and different types (cost, benefit and etc.), studying the relation between error functions and distinguishing rules when error-eliminating decision making, giving the concepts of error loss value and error limit loss value, giving the way transforming different data type attributes into error loss value, a method of hybrid multi-attributes error-eliminating decision making under uncertain condition is proposed. The method uniforms attributes value of different data types to error loss value and compares the error loss value of different strategies to determine which strategy is the best. And it’s applied to the site selection decision-making of emergency material pool.(4)The weights of experts and attributes should be determined when making group decision under uncertain conditions. Moreover current group decision-making methods usually ignore avoiding the loss when making decision wrongly. Thus, to solve the group decision-making problem from view of error-eliminating, giving the way of transforming the group decision matrix to group error loss matrix, giving the concepts of error entropy and error entropy weight and the way of determining the weights of experts and attributes with them, giving the way to determine the ideal points, giving the way to aggregate group preference using error loss value, the weighted addition method and the ideal points method of group decision-making based on error-eliminating theory are proposed in the thesis. The two methods give ways to make group decision from the perspective of avoiding error loss when making decision wrongly. An example of making decision by these two methods is also proposed in the thesis.
Keywords/Search Tags:decision-making, forecast, uncertainty, error-eliminating theory, entroopy, Support Vector Regression
PDF Full Text Request
Related items