Adaptive Depression Diagnosis Using an Improved Support Vector Machine

Abstract

The aim of this research is to modify the traditional support vector which is not adaptive to be adaptive. In the traditional SVM algorithm, the computation of the optimal plane is based only on the closet data points (support vector). In this research work, the existing support vector was modified such that instead of relying just on the closet support vector, the average distance of all the vectors (data objects) is obtained and used to compute the optimal plane. This accommodates a data point that represents a new fact or could contribute significantly which is not close to the partitioning planes to be used in construct the optimal plane. The new support vector formulation is adaptive because new facts are used to update the knowledge of the SVM training data before using it to construct a hyper-plane. The mathematical formulation showed that the new method of constructing a hyper-plane is adaptive and can accommodate new facts. Thus, this takes care of the identified lapse in traditional Support Vector. In this research work, a Support Vector that is adaptive to a new fact is formulated. This enables Support Vector to be used in an environment that is dynamic and susceptible to quick changes.