Apart from improving the performance of KCV method classification procedure by boosting parameters, it can be further improved by iteratively adjust this classifier in favor of those instances misclassified by previous iterations.
Our proposed method has many advantages as follows: (i) it handles non-linearity in a disciplined manner (ii) by introducing the pair wise class discriminant information and simultaneously employing boosting to robustly adjust the information, it effectively overcomes the problem in KCV (iii) it constitutes a strong ensemble based KCV framework having advantages of both the boosting and KCV techniques.
In this section we describe the KCV method proposed in (18) briefly.
In the above KCV method, the discriminant criterion considers only the class scatter and it neglects the neighboring class's inferences.
In KCV method each class space is modeled as separate subspaces.
These enhanced scatter operators are used in KCV method to compute the common vector in our proposed method.