Method for multi-layer classifier

A method for multi-layer classifier, applying on a computer readable medium for classifying multiple image samples including a processor, an input device, and a storage device, the method at least comprising the following steps:

• (a) receiving a plurality of samples;
• (b) providing a plurality of attributes, and evaluating a significance of the samples to the attributes by a selection criterion;
• (c) selecting at least one cut-point to establish a discriminant analysis model, which is established by providing a criterion to determine the at least one cut-point in one of the samples which is significant according to the step (b), and classifying the samples into at least one class in a current layer, wherein the at least one class comprises a first class (NodeA), a second class (NodeB), and an undefined third class (NodeN);
• (d) proceeding a step of evaluating a performance of the discriminant analysis model when adding the attributes into the discriminant analysis model, wherein the discriminant analysis model further classifies the undefined class (NodeN) and the sample thereof into a next layer when the performance of the discriminant analysis model is improved by adding the attributes, and the next layer of the discriminant analysis model is established by providing the criterion to determine the at least one cut-point of the samples which is significant according to the added attributes, and then classifying the samples into the first class (NodeA), second class (NodeB), and undefined third class (NodeN);
• and
• (e) providing a stop criterion, in which the discriminant analysis model stops classifying into the next layer when there is no significant attributes can be found at next layer;
• or in which the discriminant analysis model stops classifying into the next layer when proceeding the step of evaluating the performance of the discriminant analysis model cannot be improved by adding the attributes to the discriminant analysis model under a condition of a rejected null hypothesis.