Proposing stochastic probability-based math model and algorithms utilizing social networking and academic data for good fit students prediction

Abstract

The research progress presented in this paper comes under the areas of data science. The authors propose enhanced machine learning (supervised learning) framework for the prediction of the students through stochastic probability-based math constructs/model and an algorithm [Good Fit Student (GFS)], along with the enhanced quantification of target variables and algorithmic metrics. Academia in today’s modern world sees the problem of dropouts, low retention, poor student performances, lack of motivation, and unnecessary change of study majors and re-admissions. The authors consider this challenge as a research problem and attempt to solve it by utilizing social networking-based personality traits, relevant data and features to improve the predictive modeling approach. The authors recognize that admission choices are often governed by family trends, affordability, basic motivation, market trends, and natural instincts. However, natural gifts and talents are minimally used to select such directions in the academics. The authors based on literature review identify this a research gap and improves with a unique blend of algorithms/methods, an improved modeling of performance metrics, built upon cross-validation to improve fitness, and enhance the process of feature engineering and tuning for reduced errors and optimum fitness, at the end. The authors present the latest progress of their research in this paper. The included results show the progress of the work and ongoing improvements. The authors use machine learning techniques, Microsoft SQL Server, Excel data mining, R and Python to develop and test their model. The authors provide related work and conclude with final remarks and future work.