Achieving Personalized Precision Education Using the Catboost Model during the COVID-19 Lockdown Period in Pakistan

Abstract

With the emergence of the COVID-19 pandemic, access to physical education on campus became difficult for everyone. Therefore, students and universities have been compelled to transition from in-person to online education. During this pandemic, online education, the use of unfamiliar digital learning tools, the lack of internet access, and the communication barriers between teachers and students made precision education more difficult. Customizing models from previous studies that only consider a single course in order to make a prediction reduces the predictive power of the model because it only considers a small subset of the attributes of each possible course. Due to a lack of data for each course, overfitting often occurs. It is challenging to obtain a comprehensive understanding of the student’s participation during the semester system or in a broader context. In this paper, a model that is flexible and more generalizable is developed to address these issues. This model resolves the problem of generalized models and overfitting by using a large number of responses from college and university students as a dataset that considered a broader range of attributes, regardless of course differences. CatBoost, an advanced type of gradient boosting algorithm, was used to conduct this research, and enabled the developed model to perform effectively and produce accurate results. The model achieved a 96.8% degree of accuracy. Finally, a comparison was made with other related work to demonstrate the concept, and the experimental results proved that the Catboost model is a viable, accurate predictor of students’ performance.

DHM-Ocr

“…The presented model may be used for course recommendation and mapped to a recommendation system issue. Altaf et al [1] presented various distinct models for course performance prediction. The methodology developed in this research is broad enough to produce accurate predictions based on many inputs.…”

“…These platforms offer hundreds of courses to learners. Parallel research results on online education platforms indicate that learners face great difficulty selecting an appropriate course due to the many online course providers and the large set of course data [1].…”

Machine Learning Assessment of Damage Grade for Post-Earthquake Buildings: A Three-Stage Approach Directly Handling Categorical Features

“…CatBoost incorporates all the settings needed to build separate decision trees and set up the random forest. The literature [41][42][43] can be referenced for further information about CatBoost.…”

Research on Precision Education Management Strategies for College Students Based on the Background of Informatization Innovation and Entrepreneurship

“…How to analyze and mine the massive data of college students' behaviors and habits, and improve the precise education for college students according to their needs is one of the important problems faced by colleges and universities at present [8][9]. Asad, R et al argued that educational data mining can help teachers to better understand students' learning, and for online education, pushing accurate education to students has become the goal to be achieved by popular media education and by comparing five different machine learning methods, it was concluded that the superiority of hybrid integrated learning models in predicting students' learning outcomes [10]. Aykan, E et al took Elsies University of students to investigate the influence of personality traits on innovation and entrepreneurship of university students and concluded that there is a positive correlation between personality traits and sub-dimensions and entrepreneurial intentions, while learning is one of the most important factors affecting the personality traits of students [11].…”