Betterment of Classroom and Examination Performance Using Historical Data - A Survey

In higher education, understanding, modelling, and forecasting student performance presents considerable hurdles in terms of developing accurate and effective diagnostic models. Several researchers have attempted to construct intelligent classifiers for predicting student accomplishment, but they failed to recognise the necessity of identifying the main components that contribute to the obtained performance. Such recognition is vital for programme directors to evaluate the strengths and limitations of their academic programmes and, as a result, make the required corrections to improve student accomplishment. Consequently, there is room to present a model that improves the accuracy of predicting student academic performance, as measured by future grades in various courses, and, also an optimized multi-label classifier that predicts qualitative values for the influence of various factors on the obtained student performance, primarily using a feed-forward neural network. The goal of the experimental findings are to demonstrate significant gains over single baseline models (e.g., linear regression, matrix factorization), indicating the applicability of the suggested technique in identifying several factors influencing student performance. This study underlines the need of predicting student achievement of learning outcomes in future studies. Along with this, another model that proved to be precise and accurate in predicting the outcome in a similar but not identical scenario, was the KNN classifier, which used data to classify the graduation chances of students. Using the classifiers in the data sets, there are several analyses that can be covered in terms of question pattern, format, and more. This study aims to push a new service that can be established in several organizations and locations to benefit educational professionals and students. The datasets are naturally created in the course of the university’s classes. This allows for a lot of flexibility in terms of output variables and classifiers. Along with this, using PyCharm or another Python based IDE and Python’s built in packages and libraries to perform the analyses and build the neural models. This allows for widespread application and use.