Paper Title
Support Vector Based Approximation For Detection of Defect Forecasting Severity

There is non-linear relationship between defects and the software metrics, where complex mapping is being resulted. Therefore, focusing towards defect density, it should be like business requirement of effective and practical approach, which finds the defect density especially in pre-software releases. Soft computing which is a type of evolutionary technique provides a better platform to solve this non-linear and complexity problems. Aim of this proposed technique is to evaluate and also validate a machine learning (ML) approach in prediction of detection of defects. Important constraint of a benchmarking machine learning strategy is to define objective function based on productive universal approximation. Polynet which is based on polynomial machine learning model is used, where it is not similar to the traditional models of machine learning that are based on complexes kernel. It is also framing with the simplified kernel and its objective function is with significant universal approximation. This model is specific to the Industrial Data. Motivation which is given by polynet, redefines kernel strategy and its objective function with universal approximation, where it defines a machine learning model for finding the defect density in pre-software releases. Here support vector machine (SVM) a learning model can also be devised, which uses its objective function with universal approximation for detection of defect severity. Keywords— Universal approximation, polynomial network, support vector machine.