It is also important to choose vali

It is also important to choose validation method (classifier accuracy) for the chosen classifier. This method is known as cross-validation in which the input data is split into training and testing sets and the test set (unseen by the method or classifier) is validated against the training set to check if the classifier can reproduce the known output. Several types of cross-validation are used in literature. One of the simplest methods is the 2-fold cross-validation wherein the data is randomly split into training and test sets. An extension of the 2-fold cross-validation is the K-fold cross-validation in which the data is randomly split into K subsets. For training K À 1 subsets are chosen and the remaining subset is used for testing. This process is repeated until all the subsets are used for testing. Another version of the K-fold cross-validation is the Leave-one out
cross-validation (LOOCV) wherein K is equal to the number of samples i.e. each sample is used for testing and the rest of the samples are used for training. This process is done until all the samples are tested. For model selection (feature selection or classifier) the training set may be further split into training and validation sets. The prediction of the validation set is used to
reinforce the model selection. In [34] the authors address the problem of over fitting found when comparisons of the feature selections are made. The authors argue the cross-validation methods can contribute to over estimating the model performance. The author tests the SFS and SFFS method on various datasets with k-NN [18,6] classifier as the wrapper. They use the 2-fold and LOOCV cross-validation methods to compare the results of the two feature selection algorithms.

Kết quả (Việt) 1:

It is also important to choose validation method (classifier accuracy) for the chosen classifier. This method is known as cross-validation in which the input data is split into training and testing sets and the test set (unseen by the method or classifier) is validated against the training set to check if the classifier can reproduce the known output. Several types of cross-validation are used in literature. One of the simplest methods is the 2-fold cross-validation wherein the data is randomly split into training and test sets. An extension of the 2-fold cross-validation is the K-fold cross-validation in which the data is randomly split into K subsets. For training K À 1 subsets are chosen and the remaining subset is used for testing. This process is repeated until all the subsets are used for testing. Another version of the K-fold cross-validation is the Leave-one outcross-validation (LOOCV) wherein K is equal to the number of samples i.e. each sample is used for testing and the rest of the samples are used for training. This process is done until all the samples are tested. For model selection (feature selection or classifier) the training set may be further split into training and validation sets. The prediction of the validation set is used tocủng cố việc lựa chọn mô hình. Năm [34] các tác giả giải quyết vấn đề của hơn phù hợp tìm thấy khi so sánh các lựa chọn tính năng được thực hiện. Các tác giả cho các phương pháp xác nhận đường có thể góp phần trên ước tính hiệu suất mô hình. Tác giả kiểm tra phương pháp SFS và SFFS trên datasets khác nhau với k-NN [18,6] loại như wrapper. Họ sử dụng các 2-fold và LOOCV xác nhận qua phương pháp để so sánh kết quả của hai tính năng lựa chọn thuật toán.