学生成绩等级预测模型研究

Research on Student Achievement Level Prediction Model

摘要: 学生成绩等级的精准预测是优化学生管理、提升教学指导效能的关键支撑，本研究以Kalboard 360数据集为基础，聚焦学生在线学习平台产生的过程性数据——既涵盖学历背景等基础特征，也包含课堂举手、学习资源访问频次等行为特征(此类数据能够有效反映学生的知识掌握水平)，借助PySpark生态工具链开展建模工作：首先通过pyspark.sql库将原始数据转换为DataFrame格式，并完成数据编码等预处理；随后基于pyspark.ml库中的分类算法，分别构建逻辑回归与随机森林两类学生成绩等级预测模型，经混淆矩阵、ROC曲线等性能指标验证，随机森林模型的预测精度显著优于逻辑回归模型，而该研究的核心价值在于，模型输出结果可支撑面向学生的个性化学习建议制定，同时帮助教师及时识别学生的学习难点与问题，进而实施针对性的教学调整与指导，其应用也有助于推动教育领域智能化、个性化教学模式的落地，最终助力学生的全面发展。

Abstract: Accurate prediction of students’ academic performance grades serves as a pivotal underpinning for optimizing student management and enhancing the efficacy of teaching guidance. Based on the Kalboard 360 dataset, this study focuses on the process-oriented data generated by students on online learning platforms, including both basic attributes such as educational background and behavioral features like in-class hand-raising and learning resource access frequency. Such data can effectively reflect students’ level of knowledge mastery. This research leverages the PySpark ecosystem toolkit for model construction. First, the pyspark.sql library is used to convert raw data into DataFrame format and complete preprocessing procedures such as data encoding. Subsequently, based on the classification algorithms in the pyspark.ml library, two types of student academic performance grade prediction models (logistic regression and random forest) are constructed respectively. Verified by performance metrics including confusion matrices and ROC curves, the random forest model demonstrates significantly higher prediction accuracy than the logistic regression model. The core value of this study lies in the fact that the model output can support the formulation of personalized learning recommendations for students, while also helping teachers promptly identify students’ learning difficulties and problems, thereby implementing targeted teaching adjustments and guidance. The application of this model is conducive to promoting the implementation of intelligent and personalized teaching models in the field of education, ultimately facilitating the all-round development of students.