A New Algorithm For Conformity And Its Application

Converting multiple evaluation index to reflect the overall characteristics of the evaluation object by choosing corresponding computation form for different assessment purposes, is multi-index comprehensive evaluation. At present, comprehensive evaluation has been more widely used in the daily life and scientific research with the growing in the field of study and increasingly complex evaluation objects. However, many practice showed that all the results of comprehensive evaluation were not satisfactory. Various results may be calculated by different evaluation methods, including Analytic Hierarchy Process, Fuzzy Comprehensive Evaluation, Grey Relational Analysis, Technique for Order Preference by Similarity to Ideal Solution, Principal Component Analysis and so on. They were widely used because of their own advantages, however, two shortcomings exist which cannot be ignored:The first problem is data standardization. Generally, evaluation index have different dimension and order of magnitude, we should be ensured that data is consistent and the dimensionless processing before assessment. But different data processing methods tend to give different results because there is no uniform and reliable principle.The second problem is the difficulty of determining the weight coefficient objectively. We should set different weight coefficient of index to reflect the important degree of each index in the evaluation system, different coefficient weights can lead to very different or even contrary conclusions of evaluation. The method of expert estimate weight based on their experience may be reasonable while it is too subjective, and amateurs are difficult to grasp it; the other method of determining by the correlation between indicators or coefficient of variation often cannot accurately reflect the objective reality as it is.The thesis put forwards a new algorithm, named conformity, basing on summarizing and analyzing several comprehensive evaluation methods which are commonly used. This new algorithm uses the original data of evaluation information to calculate, shows the closeness between evaluation object and target value by the Mahalanobis distance, providing an objective and reasonable judgment of the object. In this study, first of all, confirming the relationship between index number, distance, and conformity with simulation experiments in MATLAB. Then the curve and surface fitting method was used to establish the relationship between conformity (r) and p(index number), as well as d (Mahalanobis distance).Then, taking a large number of samples drawn for testing the model, and the frequency distribution of conformity shows very good correspond to the original setting. The result fully proved the feasibility and reliability of the model. Finally, the algorithm of conformity was applied into calculating the closeness of various RVA in YANGMAI varieties and the assessment of simulated regression method. This study shows that the algorithm of conformity counts the raw data directly instead of data standardization, which greatly reduced the workload and the phenomenon of different evaluation result due to the data processing method. And it also does not need to give weight, ruled out the subjective influence, and guarantee the integrity of information and the reliability of evaluation results with comprehensive consideration of information.