GRADE ADJUSTMENTS FOR PREDICTION OF ACADEMIC PERFORMANCE: A REVIEW

A machine learning prediction of academic performance of secondary school students using radial basis function neural network

BACKGROUND

Predictive models for academic performance forecasting have been a useful tool in the improvement of the administrative, counseling and instructional personnel of academic institutions.

AIM

The aim of this work is to develop a Radial Basis Function Neural Network for prediction of students' performance using their past academic records as well as their cognitive and psychomotor abilities.

METHODS

We obtained data from a secondary school repository containing academic, cognitive and psychomotor scores of the students. The preprocessed dataset was used to train the RBFNN model. The impact of Principal Component Analysis on the model performance was also measured.

RESULTS

The results gave a sensitivity (pass prediction) of 93.49%, specificity (failure prediction) of 75%, overall accuracy of 86.59% and an AUC score (aggregate measure of performance across the possible classification thresholds) of 94%.

CONCLUSION

We established in this study that psychomotor and cognitive abilities also predict students' performance. This study helps students, parents and teachers to get a projection of academic success even before sitting for the examination.

Standardized ability tests and testing. Major issues and the validity of current criticisms of tests are discussed

At the outset a distinction was made between criticisms directed at the validity of tests and criticisms not affected by the validity of the tests. It was noted further that all criticisms of tests must take into consideration the type of test and the use to which the test is put. Criticisms of the validity of tests involved the following issues: (i) tests may be unfair to certain groups and individuals, including the extremely gifted, the culturally disadvantaged, and those who lack experience in taking tests; (ii) tests are not perfect predictors of subsequent performance; (iii) tests may be used in overly rigid ways; (iv) tests may not measure inherent qualities of individuals; and (v) tests may contribute to their own predictive validity by serving as self-fulfilling prophecies. Criticisms that are more or less independent of test validity included the effects of tests on (i) thinking patterns of those tested frequently; (ii) school curricula; (iii) self-image, motivation, and aspirations; (iv) groups using tests as a criterion for selection or allocation, or both; and (v) privacy. Several concluding remarks are in order: 1) This paper has focused almost entirely on criticisms of tests. However, the positive value of standardized tests should not be ignored. Here we must keep in mind what possible alternative measures would be used if standardized tests were abandoned. 2) We must begin thinking about tests in a much broader perspective- one that includes consideration of the social effects of tests as well as their validity and reliability. 3) Finally, an effort should be made to develop rational and systematic policies on the use of tests with the culturally disadvantaged, the dissemination of test results, and the problem of invasion of privacy. Such policies can be formulated only if we are willing to take a long hard look at the role we want testing to play in the society. Standardized tests currently are a cornerstone in the edifice of stratification in American society. It is up to the social scientist to conduct research that will enable policy makers in education, business and industry, and government to determine in a consistent and rational way the ultimate shape of this edifice.