论文部分内容阅读
目的：探讨KNN（k-Nearest Neighbor algorithm，K近邻结点算法）分类器在新疆哈萨克族食管癌分型中的应用。方法：采用KNN分类器，依据食管癌的灰度-梯度共生矩阵和灰度共生矩阵特征值，对其分型。选取样本量的40%、50%、60%作为三个训练集，K取1-29，训练并得到最优K值；选取样本量的10%到100%（以10%递增率）10个测试集，验证结果；获得最佳KNN分类模型，对模型进行评估。结果：训练结果：当K=1时，三种食管癌都能获得最高分类准确率。测试结果：改变测试集大小，当验证数据量增大时，分类准确率随之增加。最佳KNN分类模型评估：该KNN分类模型有一定的准确度，可以得到可靠的分类结果。结论：KNN分类器为新疆哈萨克族食管癌分型提供一定的依据，也为新疆哈萨克族食管癌的计算机辅助诊断系统的研发奠定基础。“,”Objective:This work shows a detailed description of the KNN(k-Nearest Neighbor algorithm) classifier when using it for Kazakh esophageal cancer classification. Methods:Using MATLB to preprocess and extract features based on gray gradient co-occurrence matrix and gray level co-occurrence matrix (GLCM) texture features,Using KNN classifier Classification on image features. First, selecting 40%、50%、60%of total sample data as training sets, We set the number of neighbors k from 1 to 29,training and getting optimal k value;Second,using ten different sizes as the classification set,which set from 10%to 90%of data(in increments of 10%) ,validating the optimal k value;Third,evaluating the best KNN classification model. Results Training Results: the optimal k value is K=1;Classification Results: accuracy with the increasing of the percentage of the data used for Classification increasing;Model evaluation: KNN classification model has a certain degree of precision and reliable. Conclusion:KNN classifier can improve the classification ability and provide a certain basis judgment of Kazakh esophageal cancer classification,this laid the foundation of computer diagnosis system for Kazakh esophageal cancer.