Post Graduate Students’ Performance Data in Competitive Examinations

Abstract

There are huge amount of data available with regard to competitive examinations for admission into Post Graduate Courses in the state of Karnataka. Typically the entrance examination for admission to the MCA course is based on Multiple Choice Question (MCQ) formats. The current research work is aimed at examining the suitability of the MCQ based formats in deriving information on the measurement of aptitude of the students for the course. The research involves analyzing students’ performance data in competitive examinations. A novel attempt in this work is to establish the linkages between past performance results and the question paper as an instrument to measure the aptitude of the students. In order to analyze the student’s performance, the power of the artificial neural networks is being attempted. Artificial neural networks(ANN)  simulate the ability of the human brain to perceive underlying patterns in a given data set not perceptible using several other well-known statistical data analysis techniques. ANN  have a set of inputs and outputs, the outputs being calculated through many iterations of transformations carried through a set of middle layers ,called as hidden layers. The number of these hidden layers can vary depending on the problem that is being tackled. For the transformation between the inputs and the outputs, the hidden layers of ANN use many triggering functions. In the current work, the students’ results, taken from the public domain website are used as the data set. These are analyzed by using ANN technique. The output from the analysis helps in the formation of clusters based on the marks obtained by the students in the competitive examination. In the next step the relationship between the marks obtained by the students and the type of questions is established. The MCQ based questions are graded using the standard benchmark as in the Bloom’s taxonomy. The ANN model can also help in predicting the results of the students. The output from the ANN model is then fed into an RDBMS platform for establishing relationships. Here, research concepts from data mining tools, clustering and document comparison have been used. The CART algorithm is then used to identify the clusters and groups of students based upon their performance and the questions that they have attempted. The computations in this research use the R programming environment for analysis using ANN and CART. The research study yielded the results in the formation of students clusters –Further research yielded the information on the type of the questions and benchmarking assigned to the questions. It was also possible for building association of the students’ cluster with the questions types attempted. The insights gained helped in obtaining suggestions for modification and improvement of the questions types, so as to develop better measures of aptitudes.