DESIGN AND IMPLEMENTATION OF A PREDICTIVE MODEL FOR STUDENT'S ACADEMIC PERFORMANCE USING MACHINE LEARNING

CHAPTER ONE

INTRODUCTION

1.1       Introduction

The use of machine learning (ML) algorithms in academia has gained significant attention in recent years due to the increasing availability of educational data and advancements in ML techniques (Yagcı, 2022). Using ML algorithms to predict students’ academic performance can give valuable insights to educators, allowing them to identify at-risk students who may need additional support, modify instructional techniques, boost learning outcomes, tailor teaching approaches to specific students’ requirements, and increase student retention rates (Adnan et al., 2021). This procedure promotes the growth of the educational system at higher institutions because educators and policymakers can intervene early to prevent students from falling behind and increase their chances of success. Applying ML algorithms to predict student academic achievement can dramatically enhance educational results and give valuable insights into the aspects contributing to academic success (Alyahyan and Du¸steg¨, 2020). Therefore, it is critical to carefully assess these algorithms’ possible benefits and limitations and ensure they are appropriately utilized. Mechanisms, such as the type of ML algorithm employed, the variables analyzed, and the assessment metrics used to determine prediction accuracy, were included as part of our investigation criteria. Applying ML algorithms in education can transform how we approach teaching and learning with a qualitative or quantitative analysis, or a mix of the two, offering an overall assessment of the condition of the results (Zhai, 2021). The potential benefits of using ML algorithms to predict academic performance extend beyond individual students and can positively impact society (Waheed et al., 2020). By improving education outcomes, individuals are better equipped to contribute to the workforce and society, leading to economic growth and social development. A vast majority of the work in educational data analysis has been devoted to developing machine learning models capable of accurately predicting students’ performance in specific contexts. However, the existing body of literature often overlooks the crucial aspect of evaluating models for their ability to transcend beyond their original training settings and demonstrate robust generalizability across diverse student populations and learning environments. This oversight raises concerns about the potential biases introduced by relying solely on ‘best-performing’ models and neglecting the search for models that exhibit superior generalization capabilities. Consequently, a pressing need emerges to address this research gap by identifying and investigating the optimal machine learning model that can be a predictive tool for assessing students’ performance. This pursuit emphasizes ensuring that the identified model achieves accurate predictions and avoids any inherent bias stemming from feature selection, thereby ensuring its applicability and effectiveness across various educational contexts. This study contributes to advancing educational data analysis practices by addressing these challenges and encouraging a paradigm shift towards holistic and unbiased model evaluation and selection.

With the increased availability of data from numerous sources, including learning management systems, online platforms, and student records, ML algorithms can give significant insights into student behavior, performance, and learning patterns (Yu et al., 2020). A thorough examination of the literature on ML algorithms for forecasting students’ academic achievement may offer a complete knowledge of the various ML approaches employed, the parameters examined, and prediction accuracy (Rastrollo et al., 2020). Institutions can profit from properly anticipating student performance by concentrating on students who are more likely to perform poorly and helping them improve their performance (Batool et al., 2023). ML algorithms used to predict students’ academic achievement can give significant insights to academics, instructors, and educational policymakers (Waheed et al., 2020; Alyahyan and Du¨¸steg¨, 2020). ML algorithms may effectively predict students’ academic achievement by analyzing different academic and non-academic criteria such as previous grades, attendance records, socioeconomic background, and student behavior (Batool et al., 2023).

A growing interest has been in using ML algorithms to predict students’ academic performance, and several studies have explored the use of ML in this area, with promising results. Several schools of thought regarding using ML to predict academic performance have emerged. One school of thought focuses on using traditional statistical methods, such as regression analysis and logistic regression. These methods assume a linear relationship between the predictor and outcome variables. For example, studies by Yaacob et al. (2019) and Waheed et al. (2020) used logistic regression while El Aissaoui et al. (2020) used linear regression to predict students’ academic performance. Another school of thought revolves around using decision trees and random forests. Decision trees are hierarchical models that predict the outcome variable through binary decisions. On the other hand, random forests are an ensemble learning approach combining numerous decision trees. Vijayalakshmi and Venkatachalapathy (2019), Altabrawee et al. (2019), and Zhang et al. (2022), for example, employed decision trees and random forests to predict students’ academic performance. A third school of thought focuses on using neural networks (Baashar et al., 2022; Liu et al., 2022), which are computational models inspired by the structure and function of the human brain to predict students’ academic performance. Neural networks are beneficial when dealing with complex, non-linear relationships between variables. Hybrid approaches also combine multiple ML methods to improve prediction accuracy. For instance, a study by Francis and Babu (2019) used a hybrid approach that combined logistic regression, decision trees, and neural networks to predict academic performance based on students’ demographic information, prior academic performance, and study habits. Each approach has its strengths and weaknesses, and the choice of method depends on the research question and the nature of the data.

Numerous factors can impact a student’s academic performance, including individual factors, such as motivation and self-regulation, and environmental factors, such as socioeconomic status and school resources (de la Fuente et al., 2021). Motivation is one of the most significant factors impacting students’ academic performance. (Deci et al, 2020)  has shown that students who are intrinsically motivated to learn, meaning that they are motivated by their interest and enjoyment of the material, are more likely to perform well academically. On the other hand, extrinsically motivated students, meaning that they are motivated by external rewards such as grades or praise, may be less likely to perform well if these rewards are not provided. Another individual factor that can impact academic performance is self-regulation, which refers to the ability to manage one’s learning and behaviors. Students who can effectively regulate their learning by setting goals, monitoring their progress, and seeking help when needed are likelier to perform well academically (Feeney et al., 2023). Environmental factors can also have a significant impact on academic performance (Asvio et al., 2022). For example, students from lower socio-economic backgrounds may have less access to resources such as high-quality schools, educational materials, and extracurricular activities, which can negatively impact their academic performance. Additionally, students who attend schools with fewer resources, such as low-income schools, may be less likely to have access to experienced teachers or advanced coursework, which can also impact academic performance. Family support and involvement in education can also have an impact on student performance. (Garbacz et al, 2021)  Research has shown that students whose families are involved in their education, such as providing support and encouragement, attending parent-teacher conferences, and monitoring homework, are more likely to perform well academically. To improve student learning concentration and collaboration in response to the COVID-19 pandemic, Nyarko et al. (2023) utilize a Discrete Choice Experiment to investigate university instructors’ preferences for current teaching strategies.

1.2       Statement of the Problem

Student success is of utmost importance for educational institutions and society. Dropouts, failures, and a drop in the standard of education in higher institutions are increasing, but identifying students who are at risk of dropping out and implementing timely interventions can greatly contribute to improving graduation rates and ensuring academic success.       

1.3       Justification of Study

The justification for this study is multi-faceted, stemming from critical gaps in current educational practices and the transformative potential of machine learning:

1. Addressing Systemic Inefficiencies: The current system at The Polytechnic Ibadan for monitoring student performance is reactive and manual. It identifies problems only after examinations, when opportunities for meaningful intervention are limited. This study is justified by the urgent need to computerize and automate this process, making it proactive and data-driven. This shift will allow the institution to support students before they fail, rather than afterward.
2. Improving Student Outcomes and Retention: A primary goal of any educational institution is to ensure student success. By accurately predicting at-risk students, the proposed system enables early and targeted interventions, such as additional tutoring, counseling, or academic advising. This proactive approach is expected to directly contribute to reducing dropout rates, improving graduation rates, and enhancing overall academic standards.
3. Optimizing Resource Allocation: Academic advisors and lecturers have limited time and resources. This system will help them focus their efforts on the students who need help the most, thereby increasing the efficiency and effectiveness of student support services.
4. Contributing to Educational Data Science: While many studies have developed predictive models, a significant research gap exists concerning the generalizability of these models across different contexts. This study aims not only to build an accurate model but also to investigate which algorithm (e.g., Random Forest vs. Logistic Regression) offers the most robust and reliable predictions, contributing valuable insights to the field of educational data mining.
5. Enhancing Institutional Decision-Making: The predictive analytics generated by the system can provide departmental and institutional leaders with insights into course difficulty, curriculum effectiveness, and trends in student performance. This data-driven intelligence can inform strategic planning, curriculum review, and quality assurance processes, ultimately strengthening the institution's academic offerings.

In summary, this study is justified by its potential to transform the educational experience at The Polytechnic Ibadan by leveraging technology to foster student success, optimize institutional resources, and advance the application of machine learning in education.

1.4       Aims and Objectives of the Study

Aim

The aim of this project is to predict students’ academic performance using random forest and logistic regression algorithms.

Objectives

       i.          To collect and compile relevant academic datasets, including students’ grades, demographic information, and behavioral records from kaggle.com. Help identify the various factors that affect students’ academic performance.

     ii.          To preprocess the collected data into a format suitable for machine learning by handling missing values, normalizing variables, and encoding categorical features.

   iii.          To explore and analyze the data to identify key academic and non-academic factors that influence student performance.

   iv.          To evaluate multiple machine learning models (e.g., logistic regression and random forest) to determine the most suitable for predicting students’ academic outcomes.

1.5       Scope of the study

The goal is to design and implement an ML model that predicts student academic performance based on historical data, demographics, and behavioral factors. The data used for this research was obtained from an online database known as Kaggle, where hundreds of collections of data are available for users to explore, analyze, and utilize in various data science and machine learning applications.  The model will assist educators in identifying struggling students and improving retention rates.

1.6       Methodology

1.6.1    Objective 1: To predict students’ final GPA classified as pass or fail, given the grades of all the mandatory courses

·       Collect academic records, including students’ grades in all mandatory courses and their final GPA.

·       Preprocess the data: normalize grade values, remove missing values, encode the GPA classification into binary labels (e.g., Pass = 1, Fail = 0).

·       Use classification algorithms (Logistic Regression, Decision Tree, and Random Forest) to model the data.

·       Train the models on the preprocessed dataset.

·       Evaluate using classification metrics like Accuracy, Precision, Recall, and F1-Score.

1.6.2    Help identify the various factors that affect students’ academic performance

·       Gather additional academic and non-academic data such as demographic data, attendance, family background, motivation levels, study habits, etc.

·       Perform exploratory data analysis (EDA) to identify patterns and correlations.

·       Use feature importance techniques (e.g., Gini importance for Decision Tree/Random Forest, coefficients in Logistic Regression) to rank the impact of different variables.

·       Optionally apply dimensionality reduction techniques (e.g., PCA) to visualize key influencing factors.

1.6.3    To implement a model to predict academic performance using Decision Tree classifiers

·       Use preprocessed training data (features such as grades, attendance, demographics).

·       Implement a Decision Tree classifier using scikit-learn or a similar ML library.

·       Tune hyperparameters like maximum depth, min samples split, etc., to avoid overfitting.

·       Train the model and validate its performance using techniques such as k-fold cross-validation.

·       Use a confusion matrix and tree visualization to interpret results and explainability.

1.7       Significance of the Study

The system offers enormous benefits to the following users:

1.     Lecturers/ Academic Advisors: The prediction model will help teachers and tutors identify weak and strong students so that teachers can lay more emphasis on instructions and procedures when dealing with weak students, to enhance overall academic performance. An academic advisor can refer to the prediction results when advising students who perform poorly in their studies so that preventive measures can be taken much earlier. In addition, a lecturer can further improve his/her teaching and learning approach, as well as plan interventions and support services for weak students.

2.     Academic Performance is an important factor people consider before applying to any tertiary institution. An institution that is known for producing low-performance students is at risk of having low intakes. The need for a Prediction Performance System arises as this will help in the early prediction of weak students and help them focus on their weak areas.

3.     Parents/Guardians/partners: Results have shown that Parents/Guardians and Partners have effects on the academic performance of Students. The Study helps to analyze the influence of family background on students’ performance predictions. Attributes such as the size of family, encouragement/motivation from parents/spouses/siblings, the highest qualification of the sponsor, and other factors will help determine those factors that affect performance.

1.8       Definition of Terms

• Academic Performance: A measure of a student’s achievement in educational activities, often represented by grades, GPA, exam scores, or course completion rates.
• Classification Algorithm: A type of machine learning algorithm used to categorize student outcomes, such as predicting whether a student will pass, fail, or drop out.
• Decision Trees: A machine learning algorithm that predicts student performance by learning decision rules based on input features such as attendance, scores, and study behavior, structured like a flowchart.
• Dropout Prediction: The process of identifying students at risk of discontinuing their studies before completion, using predictive models trained on academic, behavioral, and demographic data.
• Feature Engineering: The process of creating, transforming, or selecting relevant variables (features) from raw student data to enhance the performance of predictive models.
• Grade Point Average (GPA): A standardized measure of academic achievement calculated by averaging the grades received across courses, commonly used as a performance indicator.
• K-Nearest Neighbors (KNN): A machine learning algorithm that classifies a student’s likely performance by comparing it to the 'k' most similar students in the training dataset.
• Learning Management System (LMS): A software platform used by educational institutions to deliver, track, and manage student learning activities and data, which can be used as input for predictive models.
• Linear Regression: A statistical technique used to predict continuous academic outcomes (e.g., GPA or final exam scores) based on one or more independent variables like study hours or attendance.
• Machine Learning (ML): A subset of artificial intelligence that allows systems to learn from educational data and make predictions about student outcomes without being explicitly programmed for each scenario.
• Predictive Analytics: The use of data analysis, statistical models, and machine learning to make informed predictions about students’ future academic performance or likelihood of success.
• Student Dataset: A structured collection of data related to students, including demographic information, attendance, grades, and behavioral logs, used for training and testing machine learning models.
• Support Vector Machine (SVM): A supervised learning algorithm that classifies student performance into distinct categories by finding the optimal boundary that separates different outcome groups.

Buyers has the right to create dispute within seven (7) days of purchase for 100% refund request when you experience issue with the file received. 

Dispute can only be created when you receive a corrupt file, a wrong file or irregularities in the table of contents and content of the file you received. 

ProjectShelve.com shall either provide the appropriate file within 48hrs or send refund excluding your bank transaction charges. Term and Conditions are applied.

Buyers are expected to confirm that the material you are paying for is available on our website ProjectShelve.com and you have selected the right material, you have also gone through the preliminary pages and it interests you before payment. DO NOT MAKE BANK PAYMENT IF YOUR TOPIC IS NOT ON THE WEBSITE.

In case of payment for a material not available on ProjectShelve.com, the management of ProjectShelve.com has the right to keep your money until you send a topic that is available on our website within 48 hours.

You cannot change topic after receiving material of the topic you ordered and paid for.