【摘要】 为推动新工科背景下化学类专业实验教学的智能化与前沿化转型，文章设计并实施了一项融合机器学习与定量构效关系（QSPR）的综合性、探究性本科实验——含能化合物标准生成焓的预测。实验以162个典型含能化合物为数据集，引导学生构建初始随机森林（RF）模型，用于评估分子描述符对生成焓预测的变量重要性；进一步采用变量重要性度量法（VIM）筛选关键描述符，建立简化且高效的VIM-RF预测模型。结果表明，该模型具有良好的预测精度与泛化能力，可用于含能材料热力学性质的快速估算。实验将人工智能算法与化学基本原理深度融合，不仅强化了学生对分子结构-性质关系的理解，也培养了其数据处理、模型构建与跨学科思维能力，同时渗透绿色化学与计算替代实验的理念，有助于提升学生的科研素养与创新意识，为高校化学实验课程的智能化教学改革提供可推广的实践案例。更多还原

【Abstract】 To promote the intelligent and forward-looking transformation of chemistry laboratory teaching under the context of emerging engineering education, this study presents a comprehensive and inquiry-based undergraduate experiment integrating machine learning with quantitative structure-property relationship(QSPR) for predicting the standard enthalpies of formation of energetic compounds. Using a dataset of 162 typical energetic compounds, students are guided to construct an initial random forest(RF) model to evaluate the variable importance of molecular descriptors in predicting enthalpies of formation. Subsequently, key descriptors are selected using the variable importance measure(VIM) method, based on which a simplified and efficient VIM-RF predictive model is developed. Results demonstrate that the model exhibits good prediction accuracy and generalisation capability, enabling rapid estimation of thermodynamic properties of energetic materials. By deeply integrating artificial intelligence algorithms with fundamental chemical principles, this experiment not only strengthens students’ understanding of QSPR, but also enhances their skills in data processing, model building, and interdisciplinary thinking. It further conveys the concepts of green chemistry and computational alternatives to experimental measurements, thereby fostering scientific literacy and innovation. This work provides a replicable practical case for the intelligent reform of chemistry laboratory courses in higher education.更多还原