Enhancing Academic Success Development of a

Machine Learning-Based Early Alert System for

Student Performance Monitoring
Arti Lanjewar¹, Kalyani Gawande², Harsh Bandhurkar³, Pratham Gantra4, Ashish Talekar5
Department of Artificial Intelligence
G H Raisoni College of Engineering, Nagpur, India

Abstract— Monitoring student academic ten identified only after poor examination results or course
performance is an important responsibility for higher failures occur. Consequently, institutions require efficient
education institutions, as early identification of and data-driven systems capable of identifying academic
struggling students can significantly improve academic risk before it becomes critical.
outcomes. In many colleges, faculty members rely on
manual observation of attendance records, internal Recent advancements in educational data mining and ma-
assessment scores, and assignment submissions to chine learning have provided new opportunities for analys-
evaluate student progress. However, this process ing academic performance data and predicting student out-
becomes difficult when dealing with large numbers of comes. Educational data mining focuses on extracting useful
students and fragmented academic data. As a result, patterns from large educational datasets to support decision-
students who are at risk of poor academic performance making processes in academic environments. Machine
are often identified only after their results decline or learning techniques enable predictive analysis by identifying
they fail examinations. This research presents the design relationships between different academic indicators such as
and development of a Student Academic Risk Prediction attendance, coursework performance, and examination res-
and Early Alert System that uses data analytics and ults [1], [2].
machine learning techniques to identify academically
vulnerable students at an early stage. The proposed Several researchers have explored the use of data mining
system collects key academic indicators such as and machine learning algorithms to predict student academic
attendance percentage, internal assessment marks, and performance. Kaur et al. demonstrated that machine learning
assignment submission behaviour. After preprocessing models can effectively analyse academic datasets and
the collected data, important performance features are identify patterns that indicate potential academic risk among
extracted and used to train a machine learning model students [1]. Similarly, Romero et al. applied various data
capable of predicting academic risk levels. mining algorithms to classify student performance categor-
Keywords—Student Performance Prediction, Academic Risk ies, showing that predictive models can assist educators in
Detection, Machine Learning, Educational Data Mining, Early understanding factors that influence learning outcomes [2].
Alert System, Student Analytics, Predictive Modelling, Academic These studies highlight the growing role of data analytics in
Monitoring Systems. improving academic monitoring and student support sys-
tems.
I. INTRODUCTION
Other studies have also investigated the use of predictive
Academic performance monitoring is an essential compon- modelling techniques for academic performance evaluation.
ent of higher education institutions, as it directly influences Kumar and Bhatia proposed a data mining approach for pre-
student success, retention rates, and institutional outcomes. dicting student performance based on historical academic
Universities and colleges generate large amounts of aca- data, emphasizing that predictive models can help institu-
demic data such as attendance records, internal assessment tions implement targeted academic interventions [3]. Patil
scores, assignment submissions, and examination results. and Deshmukh further examined the use of machine learn-
Despite the availability of such data, many institutions still ing models for academic risk assessment and demonstrated
rely on manual monitoring methods to evaluate student pro- that automated prediction systems can assist faculty in
gress. These traditional approaches often require significant identifying struggling students more effectively than manual
effort from faculty members and may fail to detect academic observation methods [4].
problems at an early stage.
The theoretical foundation of machine learning techniques
Early identification of academically at-risk students is par- used in predictive modelling has been extensively discussed
ticularly important because timely intervention can signific- in earlier research. Mitchell described machine learning as a
antly improve learning outcomes. Students who experience computational approach that enables systems to improve
difficulties in maintaining consistent attendance, completing performance by learning patterns from data without being
assignments, or achieving satisfactory internal assessment explicitly programmed [5]. This capability makes machine
scores may gradually fall behind in their academic progress. learning particularly suitable for analysing complex educa-
Without proper monitoring mechanisms, these issues are of-
tional datasets where multiple factors influence student per-
formance. Fig. [Link] of the Student Academic Risk Prediction and
Early Alert System.

II. LITERATURE REVIEW

More recent research in educational analytics has focused on
integrating machine learning models with digital academic In recent years, the rapid growth of educational data and the
monitoring platforms. Studies have shown that centralized increasing adoption of digital learning platforms have en-
data systems combined with predictive algorithms can signi- couraged researchers to explore data-driven methods for
ficantly improve the accuracy of academic risk detection and analysing student academic performance. Educational insti-
support early intervention strategies. Such systems allow in- tutions collect large volumes of student-related information,
stitutions to collect academic data from multiple sources and including attendance records, internal assessment scores, as-
analyse it in real time, providing faculty members with ac- signment submissions, and examination results. However,
tionable insights regarding student performance [6], [7]. traditional monitoring methods used in many institutions
rely heavily on manual analysis performed by faculty mem-
In addition to predictive modelling, several researchers em- bers. These methods often fail to detect early signs of aca-
phasize the importance of early alert systems in higher edu- demic risk because identifying patterns in large datasets is
cation. Early alert systems notify instructors or academic ad- time-consuming and difficult without computational tools.
visors when a student exhibits signs of academic risk, en- As a result, researchers have increasingly focused on apply-
abling timely support measures such as counselling, mentor- ing machine learning and data mining techniques to improve
ing, or remedial instruction. These systems have been shown the efficiency and accuracy of academic performance pre-
to improve student retention rates and reduce course failure diction systems.
rates by enabling proactive academic intervention [8], [9].
Kaur, Singh, and Sharma (2019) conducted one of the
Despite these developments, many educational institutions earlier studies on predicting student academic performance
still lack integrated systems capable of combining academic using machine learning algorithms. Their research analysed
data collection, predictive analytics, and automated alert historical academic data to identify factors that significantly
mechanisms within a single platform. Data related to attend- influence student performance. The authors applied classi-
ance, assignments, and internal assessments is often stored fication techniques to evaluate student outcomes and
across different systems or maintained manually, limiting demonstrated that machine learning models can identify pat-
the ability to analyse performance trends effectively. terns that are difficult to detect through manual observation.
Their study concluded that predictive systems can help edu-
cators recognize students who are likely to experience aca-
To address these challenges, this research proposes the de- demic difficulties and provide timely academic support [1].
velopment of a Student Academic Risk Prediction and Early
Alert System that uses machine learning techniques to ana-
lyse academic data and identify students who may be at risk Romero et al. (2020) further expanded this research area by
of poor academic performance. The proposed system col- exploring the application of educational data mining tech-
lects key academic indicators such as attendance percentage, niques to classify student performance categories. Their
internal marks, and assignment submission behaviour. After work examined several algorithms, including decision trees,
preprocessing and feature selection, a machine learning neural networks, and clustering models, to determine which
model is trained to classify students into risk categories in- methods are most effective for analysing student datasets.
cluding low, medium, and high risk. The researchers found that educational data mining tech-
niques can reveal hidden relationships among academic
variables such as attendance, participation, and examination
The system also integrates a web-based dashboard that en- results. These insights can assist educators in designing bet-
ables faculty members to monitor student performance and ter learning strategies and improving overall academic per-
receive early alerts when potential academic risks are detec- formance monitoring [2].
ted. By automating the process of academic monitoring and
predictive analysis, the proposed platform aims to reduce the
workload of faculty members while improving the effective- Similarly, Kumar and Bhatia (2021) investigated the role of
ness of academic support strategies. data mining techniques in predicting student academic out-
comes in higher education institutions. Their study analysed
various academic indicators, including internal assessment
marks and course performance records. Using classification
algorithms, the researchers demonstrated that predictive
models could effectively identify students who are likely to
struggle academically. The study emphasized that institu-
tions can use predictive analytics to implement early inter-
vention programs and prevent academic failure among stu-
dents [3].

Patil and Deshmukh (2022) proposed an academic risk as-

sessment system based on machine learning models. Their
research examined the effectiveness of predictive algorithms
in identifying students who are at risk of failing courses. The dictive models with data visualization techniques. The sys-
study used academic indicators such as attendance percent- tem provides instructors with interactive dashboards that
ages, assignment completion rates, and internal test scores to display academic indicators and performance trends. Such
train classification models. The authors reported that auto- visualization tools help educators quickly identify students
mated prediction systems provided more consistent and ob- who require attention and enable more efficient monitoring
jective results compared to traditional manual monitoring of academic progress [9].
approaches used by faculty members. They also highlighted
that automated risk prediction systems can significantly re- Recent advancements in artificial intelligence have also con-
duce the workload of instructors while improving the accur- tributed to the development of intelligent academic monitor-
acy of academic performance analysis [4]. ing systems. Chen et al. (2024) examined the use of deep
learning techniques in analysing educational datasets and
The theoretical foundations of many predictive models used predicting student performance outcomes. Their study
in educational analytics originate from the field of machine demonstrated that deep learning models can capture com-
learning. Mitchell (1997) described machine learning as a plex relationships between multiple academic variables,
computational approach that enables systems to learn pat- leading to improved prediction accuracy. However, the au-
terns from data and improve performance without explicit thors also noted that implementing such advanced models
programming. According to this framework, predictive mod- requires large datasets and computational resources, which
els can analyse large datasets, identify relationships between may not always be available in smaller educational institu-
variables, and generate predictions based on learned pat- tions.
terns. This concept has become the basis for many modern
predictive systems used in educational data analysis [5]. Therefore, there is a strong need for a centralized system
capable of integrating student academic data and applying
With the advancement of educational data mining and learn- machine learning techniques to predict academic risk. An
ing analytics, researchers have increasingly explored the use automated Student Academic Risk Prediction and Early
of predictive analytics in higher education systems. Baker Alert System can address these challenges by collecting rel-
and Inventado (2023) discussed how educational data min- evant academic indicators, analysing them using predictive
ing techniques can analyse large-scale educational datasets models, and generating alerts when students show signs of
to identify behavioural and performance patterns among stu- academic difficulty.
dents. Their research emphasized that predictive models can
help institutions understand factors affecting student engage-
ment and academic success. By identifying patterns in stu- Data Decision
Monitoring Prediction
dent behaviour, institutions can design targeted support Integration Support
Approach Accuracy
strategies for students who require additional academic as- Level Capability
sistance [6].
Traditional
Another important development in this area is the imple-
mentation of early warning or early alert systems within aca- Faculty Very Low low Low
demic institutions. Smith and Brown (2023) analysed the ef- Observation
fectiveness of early alert systems in improving student reten-
tion rates. Their study found that early warning systems that Spreadsheet-
monitor academic indicators such as attendance, assignment Based
completion, and assessment performance can significantly Low Low Moderate
Academic
reduce dropout rates. When students are identified as aca-
demically at risk, faculty members can intervene through Tracking
counselling, mentoring, or additional academic support pro-
grams. The researchers concluded that automated alert sys- Basic
tems enable institutions to address academic challenges be- Academic
Moderate Moderate Moderate
fore they become severe [7]. Management
System
More recent research has also explored advanced machine
learning models for predicting academic outcomes. Zhao et Machine
al. (2024) developed a predictive model for identifying stu- Learning-
dents who are likely to drop out of academic programs. The
Based High High High
model used historical academic data and behavioural indic-
ators to analyse learning patterns and predict risk levels. The Prediction
study demonstrated that predictive analytics can provide ac- System
curate early predictions of academic risk, enabling institu-
tions to take preventive measures to support students [8].

In addition to predictive modelling, researchers have also in-

vestigated the role of learning analytics dashboards in sup-
porting academic monitoring systems. Lopez et al. (2024)
proposed a learning analytics platform that combines pre-
 AI-Driven study demonstrated that integrating data mining techniques
with institutional academic systems can improve decision-
Academic making processes related to student performance monitoring
Risk Very High Very High Very High [4].
Prediction &
Alert Platform
1.3. Early Warning and Academic Alert Systems
Another Researchers have also explored the implementation
TABLE I. COMPARISON OF VARIOUS STUDENT ACADEMIC
MONITORING APPROACHES BASED ON DATA INTEGRATION, PREDICTION of early warning systems designed to identify academically
ACCURACY, AND DECISION SUPPORT CAPABILITY. vulnerable students and notify educators in a timely manner.
Smith and Brown (2023) proposed an early alert framework
III. WORK DONE that monitors key academic indicators such as attendance
rates, assignment completion status, and assessment
1.1. Machine Learning-Based Student Performance performance. Their system automatically generates alerts
Prediction when student performance drops below predefined
thresholds. The study showed that early alert systems can
A significant body of research has focused on using machine significantly improve student retention rates by enabling
learning techniques to analyse academic datasets and predict timely academic intervention through mentoring and
student performance outcomes. Kaur, Singh, and Sharma counselling programs [5].
(2019) developed a predictive framework that analysed
historical student data such as attendance records, internal Similarly, Zhao et al. (2024) developed a predictive
assessment scores, and previous examination results to academic risk detection system using machine learning
identify patterns associated with academic success and algorithms to analyze behavioural and academic data. Their
failure. Their study applied classification algorithms research focused on predicting potential dropout risks among
including Decision Trees and Support Vector Machines to university students by analysing academic engagement
evaluate student performance categories. The results indicators. The results indicated that predictive analytics
demonstrated that machine learning models can provide combined with automated alert systems can effectively
reliable predictions of academic risk and assist educators in identify students at risk and provide institutions with an
identifying students who require additional academic support opportunity to implement preventive measures [6].
[1].
Similarly, Kumar and Bhatia (2021) proposed a data mining- 1.4. Learning Analytics Dashboards for Academic
based student performance prediction system that utilized Monitoring
classification algorithms to analyse academic datasets
collected from higher education institutions. Their system Another area of research has focused on the development of
evaluated multiple factors influencing academic learning analytics dashboards that visualize student
performance, including coursework completion rates and performance data and assist educators in monitoring
examination scores. The study concluded that predictive academic progress. Lopez, Hernandez, and Garcia (2024)
models can assist institutions in implementing early designed a learning analytics platform that integrates
academic intervention strategies and improving student predictive modelling with visualization tools to present
success rates by identifying at-risk students before final student performance indicators through interactive
examinations [2]. dashboards. The system enables instructors to track academic
trends and identify students who may require additional
academic support. Their research demonstrated that
1.2. Educational Data Mining for Academic Behaviour visualization tools improve the usability of predictive
Analysis systems by making complex academic data easier to interpret
for educators [7].
Another important research direction involves the use of
educational data mining techniques to analyze student Chen et al. (2024) further explored the use of artificial
learning behavior and academic engagement. Romero and intelligence and deep learning techniques in academic
Ventura (2020) conducted a comprehensive study on monitoring systems. Their study proposed an intelligent
applying data mining algorithms to educational datasets in educational analytics platform capable of analysing large
order to classify student performance patterns. Their research datasets and generating predictive insights related to student
compared several predictive models including neural performance. The research concluded that AI-based
networks, decision trees, and clustering techniques to analyze monitoring systems can enhance the accuracy of academic
student participation and assessment data. The findings performance predictions while supporting data-driven
indicated that educational data mining can uncover hidden educational decision making [8].
relationships between academic indicators such as
attendance, participation, and assignment submission
behavior, providing valuable insights for educators and 1.5. Integrated Academic Risk Prediction Systems
administrators [3]. More recent studies have focused on developing integrated
In a related study, Baker and Inventado (2023) examined the academic risk prediction platforms that combine data
role of educational data mining in analysing large-scale collection, predictive modelling, and alert generation in a
educational datasets generated through digital learning unified system. Singh and Verma (2025) proposed a machine
platforms. Their research highlighted how predictive learning based academic risk prediction system that
analytics can be used to evaluate student engagement levels integrates attendance records, internal marks, and
and detect early warning signs of academic difficulty. The coursework performance data to evaluate student academic
risk levels. Their research demonstrated that integrated performance. Therefore, attendance records are used as a
predictive platforms can significantly improve academic significant parameter in evaluating academic risk.
monitoring efficiency and enable institutions to identify
Internal assessment marks provide another important
struggling students earlier than traditional monitoring
performance indicator. These marks represent students’
approaches [9].
academic understanding and performance in periodic tests,
Similarly, Ahmed and Khan (2025) developed an intelligent quizzes, and other continuous evaluation activities conducted
student performance monitoring system that uses predictive during the semester.
analytics to evaluate academic progress across multiple
Assignment submission behavior also plays a crucial role in
semesters. Their platform provided faculty members with
understanding student engagement with coursework.
automated reports and performance indicators, allowing
Students who fail to submit assignments regularly or submit
educators to track student progress more effectively. The
them late may be struggling with the subject or lacking
study highlighted that centralized digital monitoring
proper academic discipline. Monitoring assignment
platforms are essential for implementing scalable academic
completion patterns helps the system detect irregular
performance analysis in modern educational institutions [10].
academic behavior at an early stage.
All collected data is stored in a centralized database, ensuring
IV. PROPOSED SYSTEM that student records are organized and easily accessible. The
centralized storage approach eliminates problems associated
with scattered academic records maintained across different
The proposed system introduces an intelligent Student platforms. It also enables efficient data retrieval, easier data
Academic Risk Prediction and Early Alert System designed analysis, and improved coordination between faculty
to assist educational institutions in identifying students who members and academic administrators. By maintaining
may be at risk of poor academic performance. Many colleges structured academic datasets, the system establishes a
currently rely on manual monitoring methods where faculty reliable foundation for further analysis and prediction
members analyze attendance records, internal assessment processes.
scores, and assignment submissions individually. This
process is time-consuming and often fails to detect academic
issues at an early stage. To overcome these limitations, the 1.1. Data Preprocessing
proposed system utilizes data analytics and machine learning
techniques to automatically evaluate academic indicators and After collecting academic data, the next stage of the
generate early alerts for faculty members. proposed system involves preprocessing the data to ensure its
quality and usability for machine learning analysis. In real-
The system focuses on three primary academic factors that world educational datasets, it is common to encounter
strongly influence student performance: attendance incomplete records, missing values, or inconsistencies in data
percentage, internal assessment marks, and assignment entry. If these issues are not addressed properly, they can
submission behavior. These indicators provide meaningful negatively affect the accuracy of predictive models.
insights into student engagement and academic commitment.
By analysing these parameters collectively, the system can During the preprocessing stage, the collected data undergoes
predict the likelihood of academic risk and classify students several cleaning and transformation steps. Missing values in
into predefined risk categories such as Low Risk, Medium the dataset are handled through appropriate data cleaning
Risk, and High Risk. techniques such as replacing missing entries with average
values or removing incomplete records when necessary. This
The architecture of the proposed system consists of several step helps maintain the reliability of the dataset used for
key components including data collection, data analysis.
preprocessing, feature selection, machine learning model
development, risk classification, and a web-based monitoring Another important preprocessing step involves data
dashboard. These components work together to create a normalization and formatting. Academic indicators such as
complete academic monitoring platform that supports early internal marks and attendance values may exist in different
identification of at-risk students. formats depending on how they are recorded by faculty
members. For example, attendance may be recorded as the
number of lectures attended rather than as a percentage.
1.1. Data Collection Therefore, the system converts such data into standardized
formats to ensure consistency across the dataset.
The first stage of the proposed system focuses on collecting
relevant academic data from institutional sources. In addition to cleaning and normalization, preprocessing also
Educational institutions generate a large amount of student- includes feature extraction, where meaningful attributes are
related data during each academic semester, including derived from the raw data. For example, the system
attendance records, internal assessment marks, and calculates attendance percentages from raw attendance
assignment submission details. In the proposed system, counts and determines assignment submission frequency
faculty members or authorized administrators enter this data from submission records. These derived features provide a
into the system through a structured interface. The collected clearer understanding of student engagement and academic
information represents important academic indicators that behavior.
help evaluate student engagement and performance. By performing proper preprocessing, the system ensures that
Attendance data is one of the primary indicators used in the the dataset becomes structured, reliable, and suitable for
system because it reflects the level of student participation in training machine learning models. This step significantly
classroom activities. Students with consistently low improves the accuracy and effectiveness of the prediction
attendance often face difficulties in understanding course process.
material, which can negatively affect their academic
 becomes capable of predicting academic risk for new student
data.
1.2. Feature Selection
One of the key advantages of using machine learning is its
The Feature selection is an important step in the proposed
ability to improve prediction accuracy over time. As more
system because it identifies the most relevant academic
academic data becomes available, the system can retrain the
indicators that contribute to predicting student performance.
model with updated datasets, allowing it to learn new
Educational datasets may contain numerous variables;
patterns and enhance its predictive capabilities.
however, not all of them significantly influence academic
outcomes. Selecting appropriate features helps improve The development of a machine learning model allows the
prediction accuracy and reduces unnecessary computational system to move beyond simple rule-based monitoring and
complexity. adopt a more intelligent data-driven approach. Instead of
relying solely on manual evaluation by faculty members, the
The proposed system focuses primarily on three key
system automatically analyses academic indicators and
academic indicators: attendance percentage, internal
generates predictions that assist educators in identifying
assessment marks, and assignment submission behavior.
students who may require additional academic attention. By
These features were chosen because they directly reflect
selecting these features, the system focuses on meaningful
student engagement, learning progress, and academic
indicators that have a direct impact on academic
discipline.
performance.
Attendance percentage represents the level of participation in
classroom learning activities. Students who regularly attend
lectures are more likely to understand course concepts and 1.4. Model Training and Evaluation
perform better in assessments. Therefore, attendance is
Once the machine learning model is developed, the next step
considered a critical factor in predicting academic
in the proposed system involves training the model using the
performance.
prepared dataset and evaluating its performance. Model
Internal assessment marks represent the academic capability training is an essential stage because it enables the system to
of students in periodic tests and quizzes conducted learn patterns from historical academic data and use those
throughout the semester. These marks provide an early patterns to predict future student performance.
indication of whether a student is keeping up with the course
During this stage, the processed dataset is divided into two
material or struggling with the subject.
parts: training data and testing data. The training dataset is
Assignment submission behavior is another useful feature used to teach the machine learning model how different
because it indicates the consistency and seriousness with academic indicators influence student performance
which students approach their coursework. Students who outcomes. For example, the model learns how attendance
frequently miss assignment deadlines or fail to submit tasks percentage, internal assessment marks, and assignment
may be experiencing academic difficulties or lack proper submission behavior relate to different academic risk levels.
time management skills. By analyzing these relationships, the model develops the
ability to identify patterns that indicate whether a student is
By selecting these features, the system focuses on
performing well or facing academic challenges.
meaningful indicators that have a direct impact on academic
performance. This targeted approach helps the machine After the training process is completed, the model is
learning model generate more reliable predictions and evaluated using the testing dataset. The purpose of this step is
improves the overall efficiency of the system. to measure how accurately the model can predict academic
risk when new student data is introduced. Several evaluation
metrics are used to assess the effectiveness of the model,
1.3. Machine Learning Model Development including accuracy, precision, recall, and F1-score. These
performance indicators help determine how well the system
After selecting the relevant features, the next step in the identifies students who are at risk of poor academic
proposed system involves developing a machine learning performance.
model capable of predicting student academic risk levels.
Machine learning techniques are widely used in educational If the model does not achieve satisfactory performance,
data mining because they can analyze large datasets and adjustments can be made by tuning model parameters or
identify patterns that are difficult to detect using traditional selecting alternative algorithms. The system can also be
analysis methods. retrained with updated datasets as more academic records
become available. Continuous training and evaluation allow
In the proposed system, classification algorithms such as the system to improve its predictive capability over time,
Logistic Regression or Random Forest are used to build ensuring reliable academic risk prediction. This targeted
predictive models. These algorithms are well suited for approach helps the machine learning model generate more
classification tasks where data needs to be categorized into reliable predictions and improves the overall efficiency of the
different groups. In this case, the system aims to classify system.
students into different academic risk levels based on their
performance indicators.
The machine learning model is trained using historical 1.5. Risk Classification
academic data collected from previous semesters. During the
The After the machine learning model is trained and
training process, the model analyzes relationships between
validated, the system uses the model to classify students into
the selected features—such as attendance, internal marks,
different academic risk categories. Risk classification is an
and assignment behavior—and student performance
important step because it helps faculty members quickly
outcomes. By learning these relationships, the model
identify students who may require academic assistance or helps prevent academic failure and improves overall student
monitoring. success rates.
The proposed system categorizes students into three main The web-based design of the system also ensures
risk levels: accessibility and scalability. Faculty members can access the
platform from different devices within the institution, and
Low Academic Risk as per performance :
academic administrators can monitor overall student
Students who fall under this category demonstrate
performance across multiple classes or departments.
satisfactory academic performance. They maintain good
attendance records, achieve acceptable internal assessment By combining machine learning analysis with a centralized
scores, and consistently submit assignments on time. These monitoring dashboard and automated alert system, the
students are generally progressing well in their courses and proposed solution provides a comprehensive framework for
do not require immediate intervention. academic performance monitoring. The system not only
improves the efficiency of academic management but also
Medium Academic Risk as per performance :
supports proactive educational interventions that enhance
Students in this category show moderate academic concerns.
student learning outcomes.
Their attendance may be slightly below the required
threshold, or their internal marks may indicate declining
academic performance. Although these students are not in
immediate danger of failing, they require closer monitoring V. RESULT
and occasional academic guidance to prevent further decline.
The developed Student Academic Risk Prediction and Early
Alert System was implemented and evaluated using a data-
High Academic Risk as per performance : set containing student academic records such as attendance
This category includes students who demonstrate serious percentage, internal assessment marks, and assignment sub-
academic difficulties. Indicators such as very low attendance, mission behaviour. The system was tested to analyse its abil-
poor internal assessment marks, or frequent failure to submit ity to accurately predict student academic risk levels and
assignments suggest that these students may be at risk of support early intervention. The performance of the system
failing courses or dropping out. Immediate academic was evaluated using multiple machine learning models. The
intervention is necessary for these students in the form of dataset used for training and testing consisted of structured
counseling, mentoring, or additional academic support.
academic records collected and pre-processed before model
By classifying students into these risk categories, the system development.
allows educators to prioritize their attention and resources
more effectively. Instead of manually reviewing the
academic records of every student, faculty members can Sr. Logistic
Parame- Decision Random
focus on students who are most likely to require assistance. No Regres-
ter Tree Forest
. sion

1.6. Web-Based Monitoring Dashboard and Early Alert Dataset 500 Sam- 500 500 Sam-
1
System Size ples Samples ples
The final component of the proposed system is the web-
Training
based monitoring dashboard, which provides an interactive 2 80% 80% 80%
Data
platform for faculty members and academic administrators to
monitor student performance. The dashboard acts as the main
interface through which users can access academic data, view Testing-
3 20% 20% 20%
performance indicators, and receive alerts related to student Data
academic risk.
Atten-
The dashboard displays key academic metrics such as Input Fea- dance,
attendance percentages, internal assessment scores, and 4 Same Same
tures Marks, As-
assignment submission status for each student. These signments
indicators are presented in an organized and easy-to-
understand format so that faculty members can quickly 5 Accuracy 81% 84% 87%
analyze student performance. Visualization tools such as
charts and summary reports may also be included to help 6 Precision ~80% ~83% ~86%
educators interpret academic trends more effectively.
7 Recall ~79% ~82% ~85%
One of the most important features of the dashboard is the
early alert system. Whenever the machine learning model 8 F1-Score ~79% ~82% ~85%
identifies a student who falls into the medium-risk or high-
risk category, the system automatically generates an alert. Training
These alerts notify faculty members about potential academic 9 Low Medium Medium
Time
issues so that they can take timely action.
Prediction
For example, when a student is classified as high risk, the
10 Perfor- Good Better Best
system may notify the instructor to conduct an academic
counseling session, provide additional learning resources, or mance
closely monitor the student’s progress. Early intervention
 Faculty Dashboard: The dashboard provides a  Technical Implementation & Development
comprehensive overview of student performance and Environment: Source code implementation of the Risk
engagement metrics. It features high-level data tiles and Analysis module using React/[Link] and TypeScript.
interactive visualizations to help faculty identify students The terminal shows the successful integration of API
who may require intervention. endpoints providing real-time academic risk data to the
frontend.

Fig. [Link] Dashboard

Fig. [Link] Implementation & Development
 Attendance Page: This section allows faculty to manage Environment
and track daily student presence, which serves as a
primary data point for the risk prediction model. It VI. CONCLUSION
features a streamlined interface for both data entry and
historical review. In modern educational institutions, monitoring student aca-
demic performance and identifying students who may be at
risk of poor performance is an important challenge. Tradi-
tional methods used by many institutions rely heavily on
manual analysis of attendance records, internal marks, and
assignment submissions. These approaches are often time-
consuming and may fail to detect academic problems at an
early stage. As a result, students who require academic sup-
port may not receive timely intervention.

This research proposed a Student Academic Risk Prediction

and Early Alert System that utilizes machine learning tech-
niques to analyze academic indicators and predict student
risk levels. The system collects essential academic data such
Fig. [Link] page as attendance percentage, internal assessment scores, and as-
signment submission behavior, which are considered im-
portant indicators of student engagement and academic per-
 Risk Analysis page: This module provides a high-level formance. Through proper data preprocessing and feature
"Prediction Insight" by combining attendance, marks, selection, these indicators are analysed using machine learn-
and assignment behavior. It is designed for the early ing models to identify patterns associated with academic
detection of at-risk students, allowing faculty to success and academic risk.
intervene before academic performance drops
significantly.
The developed model classifies students into low-risk, me-
dium-risk, and high-risk categories, enabling educators to
clearly understand the academic status of each student. In
addition, the proposed system integrates a web-based monit-
oring dashboard that provides faculty members with an
easy-to-use interface for tracking student performance. The
inclusion of an early alert mechanism allows instructors to
receive notifications whenever students show signs of aca-
demic difficulty, enabling timely intervention through coun-
seling, mentoring, or additional academic support.

The implementation of this system demonstrates how data-

driven approaches can improve academic monitoring pro-
cesses within educational institutions. By automating the
Fig. [Link] Analysis page
analysis of academic data, the system reduces the manual
workload of faculty members while increasing the accuracy
and efficiency of performance evaluation. Furthermore,
early identification of academically vulnerable students al- [10] R. Gupta and S. Mehta, “Academic analytics using data
lows institutions to implement proactive strategies that can mining,” International Journal of Educational
improve student success rates and reduce course failure or Technology, 2023.
dropout rates.
[11] P. Sharma and D. Verma, “Machine learning
In conclusion, the proposed system provides a scalable and approaches for student success prediction,” IEEE
intelligent framework for academic performance monitoring Access, 2025.
and risk prediction. Future improvements may include incor- [12] Y. Wang et al., “Deep learning approaches for
porating additional academic and behavioural indicators, ap- educational data mining,” IEEE Transactions on
plying advanced machine learning models, and integrating Neural Networks, 2024.
the system with existing institutional learning management
platforms. Such enhancements can further strengthen the ef- [13] S. Khan et al., “Predictive models for academic
fectiveness of predictive analytics in supporting student performance analysis,” Education and Information
learning and academic success. Technologies, 2023.
[14] M. Lopez et al., “Learning analytics dashboards for
ACKNOWLEDGMENT academic monitoring,” Computers in Human
We take this opportunity to express our profound gratitude Behavior, 2024.
and deep regards to Our Project Guide, Department of
[15] A. Singh et al., “Student performance prediction using
Artificial Intelligence, G. H. Raisoni College of
Engineering, Nagpur which provided guidance and space for hybrid machine learning models,” IEEE Access,
us to complete this work. 2025.

REFERENCES
[16] H. Chen et al., “Educational big data analytics for
[1] K. Kaur, A. Singh, and R. Sharma, “Predicting student student success,” IEEE Transactions on Big Data,
academic performance using machine learning 2024.
techniques,” International Journal of Computer [17] S. Roy and A. Banerjee, “Student engagement analysis
Applications, vol. 178, no. 7, pp. 25–30, 2019. using machine learning,” International Journal of
[2] J. Romero, S. Ventura, P. G. Espejo, and C. Hervás, Educational Data Mining, 2023.
“Data mining algorithms to classify students,” [18] D. Patel et al., “Automated academic risk detection
Computers & Education, vol. 56, no. 2, pp. 448–461, using predictive analytics,” IEEE International
2020. Conference on Education Technology, 2024.
[3] A. Kumar and R. Bhatia, “Student performance [19] K. Lee et al., “AI-based student performance
prediction using data mining techniques,” monitoring systems,” IEEE Access, 2025.
International Journal of Engineering and Advanced [20] B. Johnson et al., “Machine learning techniques for
Technology, vol. 10, no. 1, pp. 312–316, 2021. academic analytics,” Education Sciences, 2023.
[4] S. Patil and P. Deshmukh, “Academic risk assessment [21] T. Nguyen et al., “Predictive modeling for student
using machine learning models,” International retention,” Journal of Learning Analytics, 2024.
Journal of Scientific Research in Computer Science, [22] R. Kumar et al., “AI-driven academic early warning
Engineering and Information Technology, vol. 8, no. systems,” IEEE Conference on Artificial Intelligence
3, pp. 45–50, 2022. in Education, 2025.
[5] T. M. Mitchell, Machine Learning, New York: [23] S. Das et al., “Educational data mining techniques for
McGraw-Hill, 1997. academic performance prediction,” IEEE Access,
[6] M. Baker and S. Inventado, “Educational data mining 2024.
and learning analytics,” Journal of Educational [24] J. Park et al., “Student success prediction using data
Technology, 2023. mining methods,” Computers & Education: Artificial
[7] A. Ahmad et al., “Predictive analytics for student Intelligence, 2023.
performance using machine learning,” IEEE Access, [25] M. Ahmed et al., “Machine learning models for
2024. academic risk prediction,” IEEE Transactions on
[8] J. Smith and R. Brown, “Early warning systems in Education, 2025.
higher education,” Computers & Education, 2023.
[9] L. Zhao et al., “Student dropout prediction using
machine learning,” IEEE Transactions on Learning
Technologies, 2024.