An AI-Powered Academic Performance Prediction System for Undergraduate Students
Notice: This is a sample project for study and reference. Submitting it as your own work violates most universities' academic integrity policies.
Abstract
About This Research Topic
By the time a university releases semester results, the window for helping a struggling student has already closed for that term. Continuous assessment scores, attendance records, and study habits all carry warning signs weeks earlier, but most institutions never systematically look at them until it's too late to act. This project takes a different approach: building an AI system that reads those signals early and hands academic advisers a ranked, explainable list of students who need attention now, not next semester. It's a strong example of the kind of applied computer science project work that combines real machine learning technique with a genuinely deployable tool, rather than stopping at an offline accuracy score.
This article walks through how the system was built, from the 220-student survey and institutional records behind it, through the four competing machine learning models tested, to the web dashboard that turns predictions into something an adviser can actually act on. It closes with what the results mean for academic advising and where a system like this can go next.
Main Abstract
Early identification of undergraduate students at risk of poor academic performance is critical for enabling timely academic intervention, yet most tertiary institutions still rely on end-of-semester results to identify struggling students, by which point corrective action is often too late to be effective. This study addresses that problem by designing, implementing, and evaluating an artificial-intelligence-powered system that predicts undergraduate academic performance from a combination of academic history, continuous assessment scores, attendance records, and self-reported study-habit and engagement factors. The study adopted the Design Science Research (DSR) methodology together with the Cross-Industry Standard Process for Data Mining (CRISP-DM) for the data-driven components of the work. Data were collected through a structured questionnaire administered to 220 undergraduate Computer Science students, covering demographic information, study habits, class attendance, and engagement indicators, combined with matching institutional academic records (CGPA and course-level grades) obtained with appropriate consent.
The combined dataset was cleaned, encoded, and used to train and compare four predictive models: Logistic Regression and Random Forest for at-risk classification (pass/at-risk), and a Random Forest Regressor and a feed-forward Artificial Neural Network (ANN) for continuous CGPA-band prediction. Models were evaluated using accuracy, precision, recall, F1-score, and confusion-matrix analysis for the classification task, and R-squared, RMSE, and MAE for the regression task. The Random Forest classifier achieved the best classification performance, with an accuracy of 89.6%, precision of 88.7%, recall of 89.1%, and an F1-score of 88.9%, outperforming Logistic Regression (81.4% accuracy). For continuous performance prediction, the ANN achieved an R-squared of 0.81 and an RMSE of 0.34 grade points, marginally outperforming the Random Forest Regressor (R-squared of 0.77). Feature-importance analysis identified continuous assessment score, class attendance rate, and weekly self-study hours as the strongest predictors of academic outcome.
The trained Random Forest classifier was embedded in a web-based dashboard that allows academic advisers to view a ranked list of at-risk students each semester, together with the key factors driving each prediction, while a complementary student-facing view allows individual students to see their own risk indicator and general improvement suggestions. System testing showed an average dashboard response time of 0.6 seconds, and a usability evaluation with academic advisers and students returned generally favourable ratings for clarity and perceived usefulness. The study concludes that ensemble machine learning models, informed by both institutional records and self-reported engagement data, can meaningfully support early identification of at-risk undergraduates, and recommends integration with existing student-information systems and periodic model retraining as the student population and curriculum evolve.
Chapter One Preview
Background to the Study
Academic performance is one of the most closely watched indicators of institutional quality in higher education, shaping everything from university rankings to individual student employability and self-esteem. Yet a significant proportion of undergraduate students in tertiary institutions struggle academically each semester, with some eventually withdrawing, being placed on probation, or graduating with results well below their potential. In many institutions, the first formal signal that a student is struggling is the release of semester examination results, by which time an entire semester of learning has already been lost and any remedial intervention can only affect future performance rather than the results already recorded.
Educational Data Mining (EDM) and Learning Analytics (LA) have emerged over the past decade as fields dedicated to extracting actionable insight from the growing volume of data that educational institutions routinely generate: continuous assessment scores, attendance logs, learning-management-system interaction logs, and increasingly, self-reported study-habit data collected through surveys. EDUCAUSE's research on learning analytics in higher education describes this as the use of data, analysis, and predictive modelling to improve teaching and learning, examining patterns and correlations that give learners, instructors, and institutions actionable insight. Machine learning techniques applied to this kind of data have demonstrated the ability to predict, often with useful accuracy, which students are likely to underperform well before final examinations, creating a genuine window of opportunity for academic advisers to intervene through counselling, tutorial referral, or adjusted study plans.
The field has matured considerably since the founding of the International Educational Data Mining Society in 2011, which now runs an annual flagship conference and a dedicated journal tracking exactly this kind of predictive work. This study contributes to that growing body of work by designing an AI-powered academic performance prediction system tailored to the undergraduate Computer Science context. The system combines institutional academic records with a structured survey of study habits and engagement factors to train machine learning models capable of flagging at-risk students early in a semester, and delivers these predictions through a web-based dashboard intended for use by academic advisers and, in a limited form, by students themselves.
Statement of the Problem
Despite the availability of continuous assessment scores, attendance records, and other data throughout a semester, most tertiary institutions do not systematically analyse this data to identify at-risk students before final results are released. Academic advising, where it exists, tends to be reactive rather than predictive, triggered by a poor result already on record rather than by early warning signs visible earlier in the semester. Advisers who wish to identify struggling students manually must sift through spreadsheets of continuous assessment scores and attendance registers, a process that is time-consuming, inconsistent across advisers, and difficult to scale across large class sizes.
While the wider literature confirms that machine learning models can predict student performance with reasonable accuracy, many published studies stop at reporting offline model accuracy and do not extend to a deployable, adviser-facing tool that translates a prediction into an actionable, interpretable early-warning signal. This study addresses that gap directly by not only training and benchmarking predictive models but also embedding the best-performing model in a usable dashboard that surfaces the key factors behind each prediction, so that advisers can act on the result rather than treating it as an opaque score.
Aim and Objectives of the Study
The aim of this study is to design, implement, and evaluate an AI-powered system that predicts undergraduate academic performance from academic and engagement data, and to deliver these predictions through an adviser-facing early-warning dashboard. The specific objectives are to:
● review existing literature and systems relevant to educational data mining and student performance prediction;
● design and administer a structured questionnaire to collect study-habit and engagement data from undergraduate students, and combine this with institutional academic records;
● preprocess and encode the combined dataset for use in classification and regression modelling;
● train and compare machine learning models (Logistic Regression and Random Forest for at-risk classification; Random Forest Regressor and Artificial Neural Network for continuous CGPA-band prediction);
● identify the features most strongly associated with academic risk using feature-importance analysis;
● implement a web-based dashboard that presents predictions and their contributing factors to academic advisers and students; and
● evaluate the overall system through model performance metrics, system response-time testing, and a usability assessment.
Research Questions
1. What machine learning techniques have been applied to student academic performance prediction in existing literature, and what factors do they typically use?
2. Which combination of academic and self-reported engagement features is most predictive of undergraduate academic risk?
3. How does a Random Forest classifier compare with Logistic Regression, and how does an Artificial Neural Network compare with a Random Forest Regressor, for predicting student academic outcomes?
4. What is the response time and usability of a web-based dashboard built around the trained models?
5. What practical value can such a system offer to academic advisers and students?
Significance of the Study
This study is significant to several stakeholder groups. To academic advisers and departments, the system offers a data-driven tool for prioritising limited advising time toward the students most likely to benefit from early intervention. To students, the system offers visibility into their own risk status and the factors driving it, potentially motivating earlier behavioural change than waiting for a poor result.
To the academic community, the study contributes an empirical comparison of classification and regression approaches to student performance prediction using a dataset that deliberately combines institutional records with self-reported engagement survey data, an approach less commonly reported in the reviewed literature. Students and researchers exploring similar machine learning and data-driven final year project ideas will find the four-model comparison structure here, two classifiers and two regressors evaluated on the same underlying dataset, a genuinely reusable template for their own work. To institutional management, the study demonstrates a low-cost, extensible approach to learning analytics that does not require a large enterprise learning-management-system deployment to begin generating useful early-warning signals.
Scope of the Study
This study is limited to undergraduate students within a Department of Computer Science and focuses on predicting two related outcomes: a binary at-risk/not-at-risk classification and a continuous CGPA-band estimate for the coming semester. Data for the study were collected through a structured questionnaire administered to 220 students, combined with matching institutional continuous assessment, attendance, and CGPA records obtained with appropriate consent and anonymisation.
The study covers the complete pipeline from data collection and preprocessing, through model design, training, and evaluation, to the implementation of a web-based dashboard and system-level testing. It does not extend to course-level grade prediction for individual courses, integration with a live Student Information System, or long-term longitudinal tracking of students across multiple academic sessions, although these are identified as directions for further work.
Operational Definition of Terms
● Educational Data Mining (EDM): the application of data mining techniques to data generated in educational settings to uncover patterns relevant to learning and academic outcomes.
● Learning Analytics (LA): the measurement, collection, analysis, and reporting of data about learners and their contexts, for purposes of understanding and optimising learning and the environments in which it occurs.
● At-Risk Student: in this study, a student whose predicted CGPA band or classification output falls below a threshold associated with a meaningfully elevated probability of academic underperformance.
● Continuous Assessment (CA) Score: the aggregate score obtained by a student from tests, assignments, and other in-semester evaluations, prior to the final examination.
● Feature Importance: a measure indicating the relative contribution of each input variable to a machine learning model's predictions.
● Early-Warning Dashboard: a web-based interface that presents predicted risk status and contributing factors to academic advisers and/or students in a timely, actionable format.
● CGPA Band: a discretised range of Cumulative Grade Point Average, such as First Class or Second Class Upper, used as a coarse-grained regression/classification target in this study, an approach consistent with how bodies like Nigeria's National Information Technology Development Agency frame applied, locally grounded AI research as a national development priority.
Conclusion
The most useful part of this project isn't the accuracy number, it's the decision to stop at a trained model. Turning a Random Forest classifier into a dashboard that an adviser can open, scan, and act on is what separates a research exercise from a tool that actually changes how a department runs academic support. The 89.6% classification accuracy and the identification of continuous assessment score, attendance, and self-study hours as the strongest predictors give any department a concrete, low-cost starting point for building its own early-warning system. For students building similar AI and data science project topics, this project is a solid template for how to move from a trained model to something genuinely usable.
Frequently Asked Questions
What is an AI-powered academic performance prediction system?
It is a system that uses machine learning models, trained on academic and engagement data such as continuous assessment scores, attendance, and study habits, to predict which students are likely to underperform before final results are released, so advisers can intervene early.
Which machine learning models were compared in this study?
Four models were trained and compared: Logistic Regression and Random Forest for at-risk classification, and a Random Forest Regressor and a feed-forward Artificial Neural Network for continuous CGPA-band prediction.
Which model performed best, and how accurate was it?
The Random Forest classifier performed best for at-risk classification, achieving 89.6% accuracy, 88.7% precision, 89.1% recall, and an 88.9% F1-score, outperforming Logistic Regression's 81.4% accuracy.
What data was used to train the models?
The study combined a structured questionnaire on study habits and engagement, administered to 220 undergraduate Computer Science students, with matching institutional records including CGPA and course-level grades, collected with appropriate consent.
What factors were found to predict academic risk most strongly?
Feature-importance analysis identified continuous assessment score, class attendance rate, and weekly self-study hours as the three strongest predictors of academic outcome.
How is the prediction delivered to academic advisers?
The trained Random Forest classifier is embedded in a web-based dashboard that gives advisers a ranked list of at-risk students each semester, along with the key factors driving each individual prediction.
Can students see their own risk prediction?
Yes. A complementary student-facing view lets individual students see their own risk indicator along with general improvement suggestions, rather than reserving the information for advisers only.
How fast is the dashboard, and was it tested for usability?
System testing showed an average dashboard response time of 0.6 seconds, and a usability evaluation with academic advisers and students returned generally favourable ratings for clarity and perceived usefulness.
What methodology guided the system's design?
The study followed Design Science Research (DSR) for the overall system-building process, combined with the Cross-Industry Standard Process for Data Mining (CRISP-DM) for the data-driven modelling components.
What are the main limitations of this system?
The dataset comes from a single department and cohort of 220 students, which limits generalisability; self-reported study-habit data carries some social-desirability bias; and the system was evaluated in a controlled testing environment rather than a full live academic-session deployment.
Purchase to unlock the full material.