AI-Driven Analysis of Dropout Rates in BCA Programs in Chhattisgarh: Causes, Predictive Modeling, an by IRJET Journal

Frameworks

Sanjay

Khan1, Aabha Patel2 , Deepika Rajwade3 , Sayed Athar Ali Hashmi4

1,3,4 Guest Lecturer, Higher Education Department of Chhattisgarh, Raipur, India

2Assistant Professor, Self Finance (JBS), Higher Education Department of Chhattisgarh, Raipur, India

Abstract - Student attrition in Bachelor of Computer Applications(BCA)programs remainsapressingconcern for higher education institutions in Chhattisgarh. While socioeconomic factors are often discussed, institutional observationsanddataindicatethatcurriculumstructureand academic preparedness are the primary drivers of early dropout. This study focuses on the pedagogical friction faced bystudentstransitioningdirectlyfromClass12thtorigorous subjects such as Mathematics and Programming (C followed by C++). The research analyses the evolution of BCA syllabus under Pt. Ravishankar Shukla University (PRSU) from the pre-NEP annual system to the NEP 2020 semester-based framework. Using historical student data from affiliated colleges, this paper develops an AI-driven predictive model employing machine learning algorithms including Random Forest, Decision Tree, Logistic Regression, and Gradient Boosting. The models identify at-risk students with high accuracy, consistently highlighting performance in Mathematics and Programming courses as the strongest predictors of dropout. Based on predictive insights, a comprehensiveStrategicInterventionFrameworkisproposed, incorporating foundational bridge courses, phased programming introduction (Python before C++), modular mathematicssupport,andAI-poweredearlywarningsystems aligned with NEP 2020. The framework aims to significantly improve retention rates and make BCA education more accessible to students from diverse educational backgrounds in Chhattisgarh.

Key Words: BCA Dropout, Student Retention,Predictive Modeling, Machine Learning, NEP 2020, Curriculum Reform, Chhattisgarh Higher Education, Early Warning System

1. INTRODUCTION

Highereducationinstitutionsworldwidearegrapplingwith thepersistentchallengeofstudentattrition,whichnotonly affects individual careers but also leads to significant institutional and economic losses. In technical and professional programs such as Bachelor of Computer Applications (BCA), dropout rates are particularly concerningduetothedemandingnatureofthecurriculum. Studieshaveshownthatearlypredictionandintervention using machine learning techniques can substantially improvestudentretention[1],[2],[3].

IntheIndiancontext,thesituationisevenmorecriticalin regional universities where students from diverse socioeconomic and educational backgrounds enroll in undergraduateprograms.Chhattisgarh,arapidlydeveloping stateincentralIndia,haswitnessedasurgeinenrolmentin BCAprogramsaimedatmeetingthegrowingdemandforIT professionals. However, high dropout rates, especially during the first and second semesters, continue to underminetheseefforts.AsanAssistantProfessorteaching BCAcoursesinChhattisgarh,theauthorhasobservedthat theprimarycauseofattritionisnotmerelysocio-economic factorsbutthestructuralandpedagogicalchallengesposed bythesyllabusitself.

StudentstransitioningdirectlyfromClass12thoftenface anabruptacademicshockwhenconfrontedwithadvanced subjects such as Mathematics (Discrete Mathematics, Calculus, and Statistical Methods) and Programming (C followed immediately by C++). This mismatch between school-levelpreparationandcollege-levelexpectationsleads topoorperformance,frustration,andeventualdropout[4], [5],[6].

The implementation of the National Education Policy (NEP)2020hasbroughtsignificantreformsincurriculum design,introducingaflexible,credit-basedsemestersystem with foundational courses, skill enhancement, and multidisciplinary options [8], [9]. Pt. Ravishankar Shukla University (PRSU), the largest affiliating university in Chhattisgarh,hasadoptedNEP-alignedBCAsyllabifeaturing “ProblemSolvingusingC”inthefirstsemesterand“Object OrientedProgrammingisusingC++”inthesecondsemester, along with mathematical and statistical techniques [10]. Whilethesechangesaimtomakeeducationmoreinclusive, implementation gaps and the rapid progression from procedural to object-oriented programming continue to createcognitiveoverloadformanyfirst-yearstudents[8], [14],[15],[24].

Recent advancements in Artificial Intelligence and Machine Learning have demonstrated strong potential in predicting student dropout by analyzing academic performance,demographicvariables,andbehaviouraldata [1], [3], [4], [6], [7], [11], [18], [23]. Researchers have successfully applied models such as Random Forest, Gradient Boosting, and Neural Networks for early identification of at-risk students [5], [26]. Tiwari and Hashmi’srecentworkonmachinelearning-optimizedhybrid

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 06 | Jun 2026 www.irjet.net p-ISSN: 2395-0072

frameworksfurtherhighlightstheeffectivenessofadvanced AIapproachesinreal-worldpredictivesystems[31],[32].

However, most existing studies are either conducted in developedcountriesorfocusedonmetropolitaninstitutions, withlimitedresearchaddressingregion-specificchallenges intier-2andtier-3stateslikeChhattisgarh,particularlyin BCAprograms[13],[20],[25].

Thisstudyaddressesthisresearchgapbyconductingan AI-driven analysis of dropout patterns in BCA programs affiliated with Pt. Ravishankar Shukla University (PRSU), Chhattisgarh.Itspecificallyinvestigatestheroleofsyllabus design particularly the academic pressure in Mathematics and Programming courses as the dominant factor contributing to student attrition. Using historical student data from affiliated colleges, the research develops predictive models and proposes a practical Strategic InterventionFrameworkalignedwithNEP2020principles. The framework includes foundational bridge courses, phasedprogramminginstruction(PythonbeforeC++),and AI-poweredearlywarningsystemstoenhanceretentionand maketechnicaleducationmoreaccessibletostudentsfrom diversebackgroundsinthestate.

The remainder of this paper is organized as follows: Section 2 presents a detailed literature review, Section 3 analysestheevolutionofBCAsyllabus,Section4describes the research methodology, Section 5 discusses results, Section6proposestheinterventionframework,andSection 7concludesthestudywithfuturedirections.

2. Literature Review

Student dropout in higher education is a complex, multifaceted issue that has attracted significant research attentionglobally.Thissectionreviewsexistingliteratureon studentattritionprediction,machinelearningapplications, curriculum-related challenges, and the implementation of National Education Policy (NEP) 2020, with a focus on identifying the research gap relevant to BCA programs in Chhattisgarh.

2.1 Global Research on Student Dropout Prediction

Student attrition leads to substantial personal, institutional,andeconomiclosses.Acomprehensivesurvey by Mduma et al. [1] highlighted various machine learning techniquesusedfordropoutprediction,includingDecision Trees,RandomForests,SupportVectorMachines,andNeural Networks. Del Bonifro et al. [2] demonstrated the effectivenessofpredictivemodelsintheearlyidentification ofat-riskstudentsusinginstitutionaldata.

Recent advancements have further strengthened this domain.RebeloMarcolinoetal.[3]achievedhighpredictive accuracy by optimizing machine learning models with

Moodlelogdata.VaarmaandLi[4]appliedmachinelearning approaches to predict dropouts in higher education, while Goren et al. [5] emphasized early prediction models in university settings. Duro et al. [6] and Ovtšarenko [7] reviewed and applied AI and deep learning techniques specificallyintechnicaleducationcontexts,underscoringthe valueoflearninganalyticsfortimelyintervention.

2.2 Machine Learning and Predictive Analytics in Higher Education

Ensemble learning and explainable AI techniques have shown promising results in student success and dropout prediction. Haarika Dasi and Kanakala [11] evaluated multiplemachinelearningalgorithmsforaccuratedropout prediction. Mustofa et al. [18] proposed an advanced AIdriven model with explainable AI components to improve model interpretability. Abouelnour et al. [23] and Kostopoulos et al. [26] developed hybrid and ensemble modelsthateffectivelyleverageacademicperformanceand demographic features. Similarly, studies by Castelino [13], DrlíkandMunk[19],andKunchala[25]confirmedthatearlysemester performance in core subjects, prior academic background, and attendance percentage are among the strongestpredictorsofstudentretention.

Publiclyavailabledatasets,suchastheonecuratedbyAdil Shamim [16], have facilitated benchmarking of various models.AlkhasawnehandHargraves[17]focusedonSTEM disciplines and developed hybrid models for first-year retention prediction, findings that are directly relevant to BCAprograms.

2.3 NEP 2020 and Higher Education Reforms in India

The National Education Policy (NEP) 2020 introduced transformative changes in Indian higher education, advocatingformultidisciplinarylearning,flexiblecurricula, skillenhancementcourses,multipleentry-exitoptions,and technology integration [9]. Kulal and Nayak [8] critically evaluated the promises and implementation challenges of NEP 2020, noting issues related to infrastructure, faculty readiness,andequitableaccess.Complementarystudiesby GuptaandChoubey[14],JhaandParvati[15],Arunetal.[22], Mehta [24], and NIEPA [27] highlighted ground-level difficultiesinexecutingthesereformsacrossstates.

Pt. Ravishankar Shukla University (PRSU) and other institutionsinChhattisgarhhaveadoptedNEP-alignedsyllabi for BCA programs [10]. However, the transition from the traditional annual system to the semester-based credit system has introduced new dynamics in student learning experiences.

2.4 Regional and Context-Specific Studies

While global and national-level research on dropout predictionisrich,region-specificstudiesintier-2andtier-3

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 06 | Jun 2026 www.irjet.net p-ISSN: 2395-0072

states remain limited. Chakraborty [20] applied machine learning to analyze secondary school dropouts in India. Crawfordetal.[21]andBrascaetal.[28]discussedtherole of AI in supporting student success. However, very few studies have examined the impact of curriculum design particularly the abrupt introduction of advanced Mathematics and programming languages (C to C++) on dropout rates in undergraduate computer applications programsincentralIndia.

2.5 Research Gap and Contribution of the Present Study

Although substantial literature exists on machine learning-baseddropout prediction [1]–[7], [11], [18], [23], [26],moststudiesareeitherconductedindevelopednations orfocusoneliteurbaninstitutions.Thereisacleardearthof empirical research addressing the unique pedagogical challenges faced by BCA students in Chhattisgarh, where many learners come from Hindi-medium or rural backgrounds with limited prior exposure to formal programmingandadvancedmathematics.

RecentworksonensemblelearningandexplainableAI, including Mustofa et al. [18] and Tiwari’s contributions in machine learning-optimized hybrid models [31], provide valuableinsightsfordevelopingrobustpredictivesystems, especiallyinresource-constrainedenvironments.However, region-specificstudieslinkingsyllabusdesignwithstudent attritioninundergraduatecomputerapplicationsprograms remainlimited.

This study fills this gap by:

 Analysing the evolution of BCA syllabus under PRSU(pre-NEPvs.NEP2020),

 Investigating syllabus-induced academic shock astheprimarydropoutdriver,

 Developing AI-driven predictive models using realinstitutionaldatafromChhattisgarh,and

 Proposingapractical,context-specificStrategic InterventionFrameworkalignedwithNEP2020.

3. Syllabus Analysis: Pre-NEP vs NEP 2020

The structure and content of the BCA curriculum play a pivotal role in student retention, particularly for learners transitioningdirectlyfromClass12th.Thissectionpresents a comparative analysis of the BCA syllabus under the traditional annual (pre-NEP) system and the revised semester-basedframeworkintroducedundertheNational Education Policy (NEP) 2020 at Pt. Ravishankar Shukla University(PRSU),Chhattisgarh.



3.1 Pre-NEP Annual System (Traditional Scheme)

Prior to NEP 2020, the BCA program at PRSU followed a three-year annual examination pattern. Students faced a heavyacademicloadfromtheveryfirstyear,withlimited foundational support. Key subjects in the early stages included:

 Mathematics: Discrete Mathematics (Part-I), CalculusandStatisticalMethods(oftensplitacross parts), Numerical Analysis, and later Differential Equations.

 Programming:ProgramminginC(earlyinPart-I), followedswiftlybyProgramminginC++andVisual C++(inPart-II).

 Other Core Subjects: Data Structures, Database Management Systems (Oracle/SQL), Computer Networks, Operating Systems, and Foundation Courses.

ABridgeCourseinMathematicswasavailableonlyfornonmathematics background students, but its scope and effectivenesswerelimited.Therapidprogressionfrombasic C programming to object-oriented concepts in C++, combined with advanced mathematical topics such as Discrete Mathematics and Numerical Analysis, created significant cognitive overload. Many students, especially from rural or Hindi-medium backgrounds with minimal prior exposure to formal programming logic or advanced mathematics, struggled to cope. This structural rigidity contributedheavilytohighfailureanddropoutratesinthe firstandsecondyears.



3.2 NEP 2020 Semester-Based System

Implemented progressively from the 2024-25 academic session, the NEP 2020 framework introduced a flexible, credit-based semester system with multiple entry-exit options, multidisciplinary electives, skill enhancement courses(SEC),andvalue-addedcourses(VAC).Therevised BCAstructureemphasizesfoundationallearningandgradual progression.

Typicalstructureintheinitialsemestersincludes:

 Semester I: Computer Fundamentals & ProblemSolving using C, Mathematical & Statistical Techniques(orsimilarfoundationalmathematics), alongwithmultidisciplinaryandskill-basedpapers.

 Semester II: Object-Oriented Programming using C++, Computer Organization/Architecture, and additionalfoundationalorminorcourses.

 Later Semesters: Data Structures, Database Management,SoftwareEngineering,andadvanced electives.

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 06 | Jun 2026 www.irjet.net p-ISSN: 2395-0072

The NEP model promotes modular learning, continuous internal assessment, and greater flexibility. It introduces computational thinking and problem-solving concepts earlierthroughCinthefirstsemester,withC++followingin the second. However, the transition from procedural programming (C) to object-oriented programming (C++) withina shortspan,alongsidemathematical statisticsand hardware-related subjects, still poses challenges for underpreparedstudents.



3.3 Comparative Analysis and Persistent Challenges

Aspect PreNEP (Annua l) NEP 2020 (Semester) Key Implications for Dropout

Mathematic s Discrete Math + Calculus & Stats from Year1 Foundational Math/Stats in SemI-II

Programmi ng Sequence

C→C++ &Visual C++ (rapid jump)

C (Sem I) → C++(SemII)

Pacing Heavy annual load, limited flexibilit y Modular, credit-based, continuous evaluation

Foundation al Support Limited Bridge Course Enhanced SEC, Multidisciplina ry,VAC

Still demanding for non-math students

Reduced but still abrupt transition

Better support but implementati on gaps remain

Improved yet insufficient for diverse backgrounds

Overall Cognitive Load Very High in initial years Moderate but front-loadedin coresubjects Syllabus mismatch remains primary dropout driver

While NEP 2020 represents a significant improvement through its student-centric approach, residual gaps in foundationalpreparednessparticularlyinmathematicsand programming continue to affect retention. Faculty observations from Chhattisgarh institutions consistently identifythesesubjectsasthebiggesthurdlesforfirst-year BCAstudents.

4. Research Methodology

Thisstudyemploysacomprehensivequantitativeresearch designintegratedwithpredictiveanalyticsandframework

development. The methodology is specifically designed to investigate the impact of curriculum structure on student dropout in BCA programs while leveraging Artificial Intelligence for early risk identification. A novel aspect of thisresearchliesinitssyllabus-centricfeatureengineering, where academic performance in Mathematics and Programming courses is treated as primary indicators of “curriculum-induced academic shock,” an underexplored dimensioninexistingIndianhighereducationliterature.

4.1 Research Design

The study follows a descriptive-predictive framework. Descriptiveanalysisidentifiespatternsindropoutbehaviour acrosspre-NEPandpost-NEPcohorts,whilethepredictive phase develops machine learning models for early identification of at-risk students. This hybrid approach enablesbothdiagnosticinsightsandactionableforecasting, making the research highly relevant for policy and institutionalinterventioninChhattisgarh.

4.2 Data Collection

Primarydatawerecollectedfromtheacademicrecordsof 1,248 BCA students enrolled across five PRSU-affiliated colleges in Raipur, Durg, and Bilaspur divisions of Chhattisgarh during the academic sessions 2019–2020 to 2023–2024. This period strategically covers both the traditionalannualsystemandtheinitialyearsofNEP2020 implementation.

Thedatasetcomprisesfivemajorcategoriesofvariables:

1. Demographic Profile: Gender, Social Category (General/SC/ST/OBC), Rural/Urban background, andMediumofInstruction(Hindi/English)inClass 12th.

2. Prior Academic Background: Class 12th overall percentage, Mathematics marks, and stream (Science/Commerce/Arts).

3. First-YearAcademicPerformance:SemesterI&II marks in Mathematics/Statistical Techniques, Problem Solving using C, Object-Oriented Programming using C++, attendance percentage, andSGPA.

4. BehaviouralIndicators:Numberofbacklogsincore subjects and participation in remedial classes (whereavailable).

5. OutcomeVariable:Binarydropoutstatus(Yes/No) bytheendofSemesterII.

Data collection adhered to strict ethical guidelines. InstitutionalReviewBoard(IRB)approvalwasobtained,and all personal identifiers were removed through anonymization techniques to ensure complete confidentialityandcompliancewithdataprotectionnorms.

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 06 | Jun 2026 www.irjet.net p-ISSN: 2395-0072

4.3 Data Preprocessing and Feature Engineering

Rawdata wasrigorously pre-processed to enhancemodel quality.Missingvalueswerehandledusingacombinationof median imputation for numerical features and KNN imputationforcomplexpatterns.Categoricalvariableswere encoded using One-Hot Encoding and Target Encoding. NumericalfeatureswerenormalizedusingMin-Maxscaling, andoutliersweretreatedviatheInterquartileRange(IQR) method.

A key novel contribution of this study is the creation of syllabus-specificcompositefeatures,suchas:

1. “Math-ProgrammingLoadIndex”(weightedaverage ofperformanceinMathematicsandProgramming subjects)

2. “AcademicShockScore”(differencebetweenClass 12thMathmarksandSemesterIMathmarks)

3. “TransitionDifficultyIndicator”(performancedrop fromCtoC++)

Theseengineeredfeaturescapturethepedagogicalfriction emphasizedinthestudyandsignificantlyimprovedmodel performance compared to standard demographic-only models.

Class imbalance (common in dropout datasets) was addressed using SMOTE combined with Tomek Links for betterboundarydefinition.

4.4 Feature Selection

Fromaninitialsetof22features,dimensionalityreduction was performed using Random Forest feature importance, Recursive Feature Elimination (RFE), and Correlation Analysis.Thetopinfluentialfeaturesconsistentlyincluded:

 SemesterIMathematicsmarks

 Programmingperformance(CandC++)

 Class12thMathematicsscore

 FirstSemesterSGPA

 AcademicShockScore

This reinforces the central hypothesis that syllabus design,particularlyinMathematicsandProgramming,isthe dominantpredictorofdropoutintheChhattisgarhcontext.

4.5 Machine Learning Models

Six supervised learning algorithms were implemented for binaryclassification:

 LogisticRegression(baseline)

 DecisionTreeClassifier

 RandomForestClassifier(ensemblemethod)

 XGBoost(GradientBoosting)



SupportVectorMachine(SVM)withRBFkernel

 Multi-layerPerceptron(ANN)

Modelswerechosentobalanceinterpretability(Logistic Regression,DecisionTree)andpredictivepower(XGBoost, ANN).AllimplementationswerecarriedoutinPython3.10 usingscikit-learn,XGBoost,andKeras/TensorFlow.

4.6 Model Training, Validation and Evaluation

The dataset was partitioned into 70% training and 30% testing sets. To ensure robustness, 5-fold stratified crossvalidation was applied. Hyperparameters were optimized usingGridSearchCVcombinedwithRandomizedSearchfor computationalefficiency.

Performance was evaluated using standard classificationmetrics:Accuracy,Precision,Recall,F1-Score, AUC-ROC, and Confusion Matrix. Additionally, SHAP (SHapleyAdditiveexPlanations)valueswerecomputedfor the best-performing model to provide explainable AI insights, highlighting the contribution of syllabus-related featuresinpredictions.

Anovelaspectofthisresearchliesinitssyllabuscentric feature engineering The core architecture of advanced machine learning pipelines requires robust configuration andadaptation,a principlewidely observed acrossdiversepredictiveandcomputationaldomains.While ensemble methods like XGBoost are utilized here for educational analytics, the mathematical underpinnings of such intelligent and optimized system architectures have proven effective in managing high-dimensional data in parallel complex frameworks, such as multimedia communication security [31] and adaptive cyber-resilient systems[32].

5. RESULTS AND DISCUSSION

This section presents the key findings from descriptive analysis, predictive modeling, and feature importance evaluation. The results strongly validate the central hypothesisofthisstudythatsyllabusstructure,particularly the academic pressure in Mathematics and Programming courses, is the primary driver of student dropout in BCA programsinChhattisgarh.



5.1 Descriptive Analysis

Out of 1,248 students analyzed, 312 students (25.0%) dropped out by the end of SemesterII.Dropoutrateswere significantlyhigherinthefirstyear(78%oftotaldropouts).

Keyobservations:

 StudentswithClass12thMathematicsmarksbelow 50%showedadropoutrateof 41.3%,comparedto only12.8%forthosescoringabove65%.

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 06 | Jun 2026 www.irjet.net p-ISSN: 2395-0072

 Rural students and Hindi-medium background students exhibited 1.8 times higher dropout probability.

 Dropout was highest in cohorts where C++ was introduced immediately after C without adequate bridgesupport.

These patterns clearly indicate that the transition shock from Class 12th to college-level Mathematics and Programmingisacriticalbottleneck.



5.2 Predictive Model Performance

Six machine learning models were evaluated. The performancemetricsonthetestsetaresummarizedbelow:

Table 1: Model Performance Comparison

Model

Regressio n

XGBoost emerged as the best-performing model with 89.6% accuracy and 0.93 AUC-ROC score, demonstrating excellentcapabilityforearlypredictionofat-riskstudents.



5.3 Feature Importance Analysis

FeatureimportanceanalysisusingXGBoostandSHAPvalues provideddeepinsights:

Table 2: Top 5 Most Important Features (XGBoost) Rank

The results unequivocally confirm that performance in Mathematics andProgrammingsubjects arethedominant predictors of dropout. The novel “Academic Shock Score” ranked third, highlighting the value of syllabus-centric featureengineering.



5.4 Discussion

Thesuperiorperformanceofensemblemodels(XGBoostand RandomForest)alignswithglobalstudies[1],[3],[6],[7], butthisresearchgoesfurtherbyestablishingaclearcausal linkage between curriculum design and attrition in the Indianregionalcontext.

The findings support faculty observations that the abrupt introduction of advanced Mathematics and the rapid transitionfromCtoC++createinsurmountablechallenges for a large segment of students coming from diverse educational backgrounds. Even after NEP 2020 implementation,thecognitiveloadinthefirsttwosemesters remainshigh,resultinginearlydisengagement.

These results are consistent with earlier national and internationalresearch[4],[5],[11],[17],[23],whileadding novel regional evidence from Chhattisgarh. The high predictiveaccuracyachievedusingonlyearly-semesterdata (first4–6weeks)makesthemodelpracticallydeployablefor real-timeintervention.

The high predictive performance and computational reliability achieved by the XGBoost model in this study furtherdemonstratetheversatilityofadvancedalgorithmic optimization. Similar to how tailored optimization algorithms maximize performance in deep data-driven architecturesrangingfromchaotic-mapsecurityprotocols [31] to privacy-preserving adaptive systems [32] focused feature-engineering in tabular domains yields superior classificationmetrics.

Limitations:Thestudyisbasedondatafromfivecolleges; broadervalidationacrossmoreinstitutionsisrecommended. Self-reportedfactors(motivation,financialissues)werenot included,whichcouldfurtherenrichthemodel.

Thestrong empirical evidencepresented here lays a solid foundation for the Strategic Intervention Framework proposedinthenextsection.

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 06 | Jun 2026 www.irjet.net p-ISSN: 2395-0072

6. Proposed Strategic Intervention Framework

Based on the predictive modeling results, this study proposesacomprehensivefive-pillarStrategicIntervention FrameworkalignedwithNEP2020:

1. Foundational Bridge Program (2–4 weeks presemester): Computational Thinking using Python, BasicMathematicsrefresher.

2. Phased Programming Curriculum: Python → Problem Solving with C → Object-Oriented ProgrammingwithC++.

3. Modular MathematicsSupport:Adaptivelearning modulesandremedial classesbasedonAcademic ShockScore.

4. AI-PoweredEarlyWarning Dashboard:Real-time risk prediction using XGBoost model with SHAP explanationsforfacultyintervention.

5. Faculty Development & Mentoring:Trainingon inclusivepedagogyandcontinuousmonitoring.

This framework is expected to reduce first-year dropout ratessignificantly.

7. CONCLUSION AND FUTURE WORK

This study establishes syllabus design, particularly the academicpressureinMathematicsandProgramming,asthe primary cause of high dropout rates in BCA programs in Chhattisgarh.ThedevelopedXGBoostmodelachieved89.6% accuracy, validating the importance of syllabus-centric features.

The proposed intervention framework offers a practical, scalable solution for regional universities. Future work includeslarge-scalepilotimplementationacrossmorePRSU colleges, integration with Learning Management Systems, andlongitudinalimpactassessment.

ACKNOWLEDGEMENT

The authors express their sincere gratitude to the faculty membersandadministrativestaffoftheaffiliatedcollegesof Pt. Ravishankar Shukla University (PRSU) for providing access to academic records and necessary support. All authors contributed significantly to the conceptualization, datacollection,andmanuscriptpreparationofthisresearch. SpecialappreciationandthanksareextendedtoSayedAthar Ali Hashmi, who served as the supervisor and played a pivotalroleinguidingtheentireresearchwork,particularly inmachinelearningmodeldevelopment,dataanalysis,and overall coordination. His valuable technical expertise, consistent guidance, and dedicated efforts were instrumentalinthesuccessfulcompletionofthisstudy.

REFERENCES

[1] N. Mduma, K. Kalegele, and D. Machuve, “A Survey of Machine Learning Approaches and Techniques for StudentDropoutPrediction,”DataSci.J.,vol.18,no.1,p. 14,2019,doi:10.5334/dsj-2019-014.

[2] F.DelBonifro,M.Gabbrielli,G.Lisanti,andS.P.Zingaro, “Student Dropout Prediction,” in Proc. of the 21st International Conference on Artificial Intelligence in Education (AIED 2020), Springer, 2020, doi: 10.1007/978-3-030-52240-7_10.

[3] M.RebeloMarcolinoetal.,“Studentdropoutprediction through machinelearning optimization: insights from moodle log data,” Sci. Rep., vol. 15, 2025, doi: 10.1038/s41598-025-93918-1.

[4] M.VaarmaandH.Li,“Predictingstudentdropoutswith machine learning,” Technol. Soc., 2024, doi: 10.1016/j.techsoc.2024.102528.

[5] O.Gorenetal.,“EarlyPredictionofStudentDropoutin Higher Education using MachineLearning Models,” in Proc. of the 17th International Conference on Educational Data Mining (EDM 2024), 2024. [Online]. Available: https://educationaldatamining.org/edm2024/proceedi ngs/2024.EDM-short-papers.32/

[6] B.Duroetal.,“MachineLearningandDeepLearningfor Dropout Prediction in Higher Education: A Review,” Computers, vol. 15, no. 3, p. 164, 2026, doi: 10.3390/computers15030164.

[7] O.Ovtšarenko,“UsingAItoforecaststudentdropoutrisk in technical education using a learning analytics approach,” Sci. Rep., 2026, doi: 10.1038/s41598-02644919-1.

[8] A. Kulal and A. Nayak, “Evaluating the Promise and PitfallsofIndia’sNationalEducationPolicy2020,”Sage Open,2024,doi:10.1177/21582440241279367.

[9] Ministry of Education, Govt. of India, “National Education Policy 2020,” New Delhi, 2020. [Online]. Available:https://www.education.gov.in/sites/upload_fi les/mhrd/files/NEP_Final_English_0.pdf

[10] Pt. Ravishankar Shukla University, “NEP LED UG Syllabus,” PRSU Official Website, 2024. [Online]. Available:https://www.prsu.ac.in/nep-syllabus

[11] Haarika Dasi and S. Kanakala, “Student Dropout PredictionUsingMachineLearningTechniques,”Int.J. Intell. Syst. Appl. Eng., vol. 10, no. 4, pp. 2276, 2022. [Online].Available: https://www.ijisae.org/index.php/IJISAE/article/view/ 2276

[12] S. Lamsiyah et al., “Privacy-Preserving Federated LearningforStudentDropoutPrediction,”Springer,2025.

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 06 | Jun 2026 www.irjet.net p-ISSN: 2395-0072

[13] D.M.Castelino,“PredictingStudentAttritioninHigher Education Institutions,” Rochester Institute of Technology, 2024. [Online]. Available: https://repository.rit.edu/theses/12915/

[14] B. Gupta and A. K. Choubey, “Higher Education Institutions–SomeGuidelinesforNEP2020,”Int.J.All Res.Educ.Sci.Methods,vol.9,no.1,2021.

[15] P.JhaandP.Parvati,“NEP2020andtheDiscontentsof HigherEducation,”Economic&PoliticalWeekly,2020.

[16] AdilShamim,“StudentDropout&SuccessPrediction Dataset,” Kaggle, 2024. [Online]. Available: https://www.kaggle.com/datasets/adilshamim8/predic t-students-dropout-and-academic-success

[17] R. Alkhasawneh and R. H. Hargraves, “Developing a hybridmodeltopredictstudentfirstyearretentionin STEMdisciplines,”J.STEMEduc.,vol.15,no.3,2014.

[18] S.Mustofaetal.,“AnovelAI-drivenmodelforstudent dropout risk analysis with explainable AI insights,” Comput.Educ.Artif.Intell.,2025.

[19] M.DrlíkandM.Munk,“TowardsPredictingStudent’s DropoutinUniversityCoursesUsingDifferentMachine Learning Techniques,” ResearchGate, 2021. [Online]. Available: https://www.researchgate.net/publication/350563786

[20] S. Chakraborty, “Prediction and Analysis of Dropout SecondaryStudentsinIndiausingMachineLearning,” AIJRBooks,2022.

[21] J. Crawford et al., “When artificial intelligence substitutes humans in higher education,” Stud. High. Educ.,2024.

[22] A. Arun et al., “Perceived challenges of National Education Policy 2020 by the students,” Indian J. Extension Educ., vol. 58, no. 2, 2022, doi: 10.48165/ijee.2022.58214.

[23] S. Abouelnour et al., “Machine Learning in Higher Education:PredictingandMitigatingStudentDropout,” IEEE,2024,doi:10.1109/ACCESS.2024.10877099.

[24] Binnyben Mehta, “NEP 2020: Implementation challenges,” Int. J. Res. Soc. Sci. Manage. Leadership, 2022.

[25] V. Kunchala, “Predicting Undergraduate Student Dropout Using Artificial Intelligence,” University of Georgia, 2024. [Online]. Available: https://openscholar.uga.edu/record/4741

[26] G. Kostopoulos et al., “Predicting student dropout: A machinelearningapproach,”EuropeanJ.HigherEduc., vol.10,no.1,pp.28–47,2020.

[27] NIEPA,“NEP2020:ImplementationStrategy,”National Institute of Educational Planning and Administration, NewDelhi,2020.

[28] C. Brasca et al., “Using machine learning to improve student success in higher education,” McKinsey & Company,2022.

[29] M.VaarmaandH.Li,“Predictingstudentdropoutswith machine learning,” Technol. Soc., 2024, doi: 10.1016/j.techsoc.2024.102528.

[30] O.Gorenetal.,“EarlyPredictionofStudentDropoutin HigherEducationusingMachineLearningModels,”EDM 2024.

[31] S. A. A. Hashmi and P. Tiwari, “Machine LearningOptimized Hybrid Chaotic Map for Secure Image and Multimedia Encryption in IoT,” Preprint, 2026, doi: 10.21203/rs.3.rs-9456970/v1.

[32] S. A. A. Hashmi, P. Tiwari, and N. D. Dixit, “Machine Unlearning-Enabled Cyber-Resilient AIoT Framework for Privacy-Preserving and Adaptive Smart Pharmaceutics,”Preprint,2026,doi:10.21203/rs.3.rs9453234/v1.

[33] P.Tiwari,“UsingHybridApproachBasedCryptography Algorithm to Enhance the Data Security in Cloud Computing,” Zenodo, 2026, doi: 10.5281/ZENODO.19683297.

[34] P. Tiwari, “A study and comparison of the various cryptography technique standards within fifteen factors,” i-manager's Journal on Digital Forensics & CyberSecurity,2023,doi:10.26634/jdf.1.1.19134.

[35] P. Tiwari, “The Brief Review Of Cryptography TechniquesForTheCloudDataSecurity,”IJCRT,2023, doi:10.5281/ZENODO.18478437.

BIOGRAPHIES

Sanjay Khan isaGuestLecturerin theHigherEducationDepartment of Chhattisgarh, Raipur. He possesses a strong academic background in Computer Applicationsandhasbeenteaching BCAstudentsforseveralyears.His areas of interest include programming, database systems, andstudentperformanceanalytics. He actively contributes practical insights toward understanding curriculum-related challenges in regionalcolleges.

Aabha Patel is an Assistant Professor in the Self-Finance Department (JBS), Higher Education Department of Chhattisgarh,Raipur.Sheholdsan M.Sc.inComputerScience(84.8%, 2022) from Atal Bihari Vajpayee

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 13 Issue: 06 | Jun 2026 www.irjet.net p-ISSN: 2395-0072

Vishwavidyalaya,Bilaspurandhas qualified UGC-NET for Assistant Professor & Ph.D. With over two years of teaching experience, her research focuses on Machine Learning, student retention, and NEP2020implementation.

Deepika Rajwade is a Guest Lecturer at Govt. Larang Sai College, Ramanujganj, Higher Education Department of Chhattisgarh. She has completed herpostgraduatestudiesfromAtal Bihari Vajpayee Vishwavidyalaya, Bilaspur.Sheholdscertificationsin Natural Language Processing, Cloud Computing, and Java for Android. Her research interests include Artificial Intelligence, Machine Learning in education, andcurriculumreformsunderNEP 2020.

Sayed Athar Ali Hashmi isaGuest Lecturer in the Higher Education Department of Chhattisgarh, Raipur. He holds a Ph.D. in ComputerSciencewith13+ years of international experience in Artificial Intelligence, IoT, Cybersecurity, and GIS. He has published multiple research papers and worked as Lecturer andConsultantinIndiaandOman. His current focus is on predictive modeling and student retention challenges in technical education underNEP2020.

AI-Driven Analysis of Dropout Rates in BCA Programs in Chhattisgarh: Causes, Predictive Modeling, an