Analysis Of Comprehensive Evaluation Method Integration Modern Effectiveness And ANN & GA

With the development of cross and integration among disciplinary researches, themodern comprehensive evaluation methods gradually diversified, they are widelyused in various fields of decision-making, Information management and evaluation.Therefore, those systematic researches on their theories and validities and proposingof more valid evaluation method have great significance in discipline and application.In this thesis, combined with many kinds of knowledge of engineering, fuzzymathematics and artificial intelligence, with an overview of the theories and methodsof the modern comprehensive evaluation methods, two parts of the research work aredone. In the one part, on the basis of summarizing their theoretical system, the factorsaffecting the validity of various methods are analyzed and researched，and meanwhile,the algorithm assessing their validity is put forward. In the other part, on the basis ofanalyzing and researching ANNCEM and GACEM, considering the problem whichthe blindness and randomness in initialization of weights and threshold of ANNCEM,and the short ability of local research and premature phenomenon of GACEM, theANN&GA integration of comprehensive evaluation method is proposed.In thesis, main research contents of the validities of the modern comprehensiveevaluation methods are outlined as follows:1. Set benchmark level sequence. In order to achieve the assessment of theirvalidities，combined with sequence number summation theory（SNST）, mode theory（MT）and weighted averaging theory（WAT）, the benchmark level sequence isdetermined on the basis of the level sequences determined by many kinds of methods.2. Assess validity. Aimed to assess the validity of a method, the benchmark levelsequence and the level sequences which determined by the method whose validityassessment is set as a goal are contacted combined with specific instance, and theelements of them are composed into a pair of level assessment materials. Theircorrelation is analyzed by using the method of the Spearman Rank correlationcoefficient (SRCC), and the validity assessment of the method which needs to beanalyzed is determined by the correlation coefficient test.The simulation effect notes that the algorithm of validity is reasonable, simpleand feasible.Main research contents of the ANN&GA integration of comprehensiveevaluation method are outlined as follows:1. Analyze the algorithms of the ANNCEM and GACEM. On the one hand, ANNCEM is analyzed and researched, and it is figured out that there are the blindnessand randomness in initialization of weights and threshold of the method. Therefore,the comprehensive evaluation net learned and trained has a certain error. On the otherhand, GACEM is analyzed, and it is figured out that although it may advance toconverge to the optimized index weights by using this method in practical application,it can optimize the weights effectively.2. Research on ANN&GA integration of comprehensive evaluation method. Onthe basis of analyzing their own advantage and disadvantage of ANNCEM andGACEM, ANN&GA integration of comprehensive evaluation method is put forwardto improve the accuracy of the comprehensive evaluation, by using the optimizationfunction of GACEM to guard and optimize the initialization of weights and thresholdof ANNCEM.On overcoming the blindness and randomness in initialization of weights andthreshold of ANNCEM, and reduction of its network error, the simulation indicatesthat ANN&GA integration of comprehensive evaluation method has a good result.