Median based conversion of SGPA into percentage by cognitive methods

The academic performance of a student is commonly evaluated with the help of Semester Grade Point Average (SGPA). But such evaluation metrics are often criticized for the non-uniformity in the procedures by which SGPA is calculated in different educational institutes and subjective nature of the system. SGPA is not standardized and reliable enough to represent the exact ability of the student for a group of subjects. Due to the difference in methodology of calculating SGPA the same cannot be converted into percentage equivalent uniformly throughout the institutes and boards. That is why; many studies were conducted for determination of a standard procedure to convert SGPA to its percentage equivalent. The present study however used the advantage of Neuro-genetic models in estimation of median percentage from the SGPA of any student within a group of students studying in the same semester by introducing the scores of SGPA relative to the median of all the SGPA. The introduction of the median component reduces the computational time and also improved the performance of the cognitive model. An eight input one output feed-forward Neuro-genetic model was prepared and compared with a Neuro-heuristic model. The results show satisfactory accuracy in the latter method. The models were validated with the common performance metrics for a deterministic selection from the available alternatives but the model predictions displayed uniform accuracy and correlation in between all the neural networks considered for the study which also supports the conversion of SGPA into percentage equivalent by application of neural network modeling concepts.