Research On Sample Cultivation Theory Of Factor Databases In Factor Space

This paper arms to study the problem of sample cultivation to solve the identification problem of classification algorithm because of the limitation of training sample in order to enrich and improve the theory of factor databases in factor space.The problems of quantitative sample cultivation are studied.A rotary outer rectangular algorithm is proposed by using geometrical configuration of sample data of factor space.Profiling the base by a rotary outer rectangular contour,gradually the intersection of contours inwards approach to approximate background base.Cultivate approximate background base and obtain comparatively perfect background base of the whole data set to achieve the purpose of reducing data sample.Quantitative dispersed data processing method for ordered qualitative nature are studied.Thus,using quantitative data to solve the problem of sample cultivation of ordered qualitative nature is realized.Numerical experiment show the effectiveness and feasibility of the method of extracting the background based and quantitative sample cultivate.The problems of qualitative sample cultivation are studied.Divide data into groups,using the cultivation rule set obtained from the initial training to test the new sample,the feedback mechanism is added,multiple training on the training set until the accuracy of the same.For error identification and unidentification of the initial testing sample,the specific rules of the regulation are given.Numerical experiment results show that,This algorithm can use the new training data in time to change the inference rules,under the premise of ensuring the accuracy of testing realize the using value of sample information effectively.