| Classify the things is the starting point that people recognize the things; and it is also an important tool that people understand the world's .so taxonomy has become the basis subject of people understand the world's.In the biological, economic, social, demographic and other fields of study, there are a large number of researches in classification. At present, there are common classification methods that base on decision tree, the Bayesian classifier, the neural network, and support vector machine, and so on.In this paper, we first introduce the relevant concept of data classification, and than In-depth study some of the most commonly used classification, but we found that the above method for solving such a classification of samples, to ensure both high accuracy and speed, appears to have less than ideal. So in this paper, we give a new method that based on core-sets to solve the classification and distinction problem, and on the basis of this method, we give classification criteria of sample and classification model in R~d space.At last it takes the classification and discrimination of fan lv flowers and the cases of breast tumor for example, the results of experiments are very well. This article reads as follows research:(1) Introduction a new reduced-order method. First, permutation and combination all the samples indicators,than construct the core-sets respectively with the combination of any 2 or 3 indicators,choose the best combination,model and experiments, classification and discrimination.and comparison the result of experiments of the two-dimensional and three-dimensional space. So in the high-dimensional space, we select some of the most effective combination of indicators of data classification in order to achieve the effect of reduced-order;(2)Present a new method of data classification and discrimination. This method is based on the principle of core-sets and different from the traditional methods. The traditional method is that find a number of similar curves to divided the sample points into several categories, but classification is very complex, this method is start the category and the Rules is core-set construction. That is, find the smallest round or a ball to divide all sample points into several categories.We can find that our method is based on simple mathematical principle and relatively easy;(3) Given the criteria of point to be tested which is outside the closure or in the cross-regional .The method in this paper has given the criteria of points to be tested which are outside the closure or in the cross-regional. It will more precisely determine the samples points to be each type. In this paper, the research results are followd, first we solve the classification problem of high-dimensional samples points with core-sets .than we use the information of known sample data to detect the type of unknown sample points.And use the idea of reduced-feature and feature selection to improve the efficiency of the algorithm.Finally,we give the concept of the points that is to be test it can be reduce the rate of misjudgement further.Experiments prove that this method has achieved good classification results and higher recognition rate, Therefore, high value applications. |
