基于机器学习的学生学习成绩预测研究

打开文本图片集

中图分类号：TP311 文献标志码：A 文章编号：2095-2945（2025）25-0034-04

Abstract： Studentacademic performance predictionisof great significance for educational decision-making and personalized learning.ThispapertakessomestudentsfromtheColegeofArtsandSciencesofNortheastAgriculturalUniversityinthe2021 and02graduatingclasesastheresearchobjects.Troughquestionnairesurveys，dataonstudents‘learning habits，sef-study， cognitivedrive，andenvironmentalfactorswerecollected.Corelationanalysiswasusedtoselette1Omostrelevantfactorsand theK-meansalgorithmwasemployedforclustering，dividing thequestionnairerespondentsinto5categories，among which categories 0 and1are more likely to have failing grades.Combining theselectedfactorswith the clustering results，a GPA predictiomodelbasedonCNNwasconstructed，incorporatingkeyfeaturessuchasself-disciplineandstudyplans.Themodel architectureincludesconvolutionallayers，polinglayers，andfullconectedlayers，andaDropoutlayerwasintroducedto enhancegeneralizationability.Understrictvalidation，themodelwasevaluatedusing MSEandMAE，demonstratingexcellent GPApredictionperformance.Theresearchresultsshowthatstudents‘academicperformanceismainlyinfluencedby1Ofactors： Whethertheyhavethehabitoforganizingnotesandreviewmaterials，howtheyspendtheirentertainmenttime，theeficiencyof atendingclasses，thesituationofformulatingstudyplans，participationincompeitions，thecompletionofstudyplanswhether theyhaveaclearplanforthefuture，whethertheyatendeveningself-studyduring non-examwees，learningself-awarenes， andwhethertheyparticipateinsecondclasroomactivities.TheConvolutionalNeuralNetworkmodelshowsgoodperfomancein predicting student behavior time series data，providing a new idea for personalized learning.

Keywords： academic performance; machine learning; K-means; CNN; performance prediction

在当今社会，高等教育机构肩负着培养高素质人才的重要使命，其教育质量直接影响国家发展与进步。（剩余3120字）