This paper targets a novel classifier merging methodology for automated and precise judgement of Alzheimer's disease. The six diverse joining rules (mean, median, product, maximum, minimum, and voting) are presented with their significance in the consolidating of classifiers with that of the individual classifiers. The approval of the proposed combination procedure is performed on benchmark ADNI dataset. The underlying emphasis uncovered the four individual classifiers out of thirteen classifiers, to be specific BayesNet, linear discriminant classifier (ldc), quadratic Bayes normal classifier (udc), and Kernel Support vector machine (KSVM) from various machine learning groups, gained the best performance values of 74.77%, 71.62%, 77.76, and 76.13% separately. The classifier merging model decorated from these four best algorithms displayed a much healthier performance with a shared mean error rate of 0.2123 in contrast to the mean error rate of 0.2493 before Ensemble. Our examinations effectively show that a classifier merging procedure produces better outcomes and orders subjects more precisely than base-level classifiers.

本文针对一种新颖的分类器合并方法，以自动，准确地判断阿尔茨海默氏病。提出了六种不同的联接规则（平均，中位数，乘积，最大值，最小值和投票），它们在将分类器与各个分类器合并时具有重要意义。建议的组合程序的批准在基准ADNI数据集上执行。重点在于揭示了13个分类器中的四个单独分类器，这些分类器分别来自各个机器学习组，分别是特定的BayesNet，线性判别分类器（ldc），二次贝叶斯正态分类器（udc）和内核支持向量机（KSVM），性能值分别为74.77％，71.62％，77.76和76.13％。由这四个最佳算法构成的分类器合并模型显示出更健康的性能，共享平均误码率为0.2123，而Ensemble之前的平均误码率为0.2493。我们的检查有效地表明，与基础级别的分类器相比，分类器合并过程可产生更好的结果并更准确地对主题进行排序。