基于机器学习的高校公共资源安全评价与科学管理

Evaluation and Scientific Management of College Public Resources Security Based on Machine Learning

为了构建作为科学研究和人才培养基地的高校的节约型校园,实现以人为本、资源节约、环境友好和生态良性循环的目标,以便快速、准确评价高校公共安全运行水平,促进公共资源得到科学的管理与利用,本研究比较了随机森林算法、层次分析法和灰色关联法的相似性,建立了随机森林-灰色关联安全评价模型和评价指标体系,对安全变量进行分类和评价。通过训练样本的机器学习,得知医疗卫生防疫的权重值最高,其后是消防措施、餐饮环境、住宿环境、巡视与处理、学习办公、校门管理、机构与制度和机动车管理,其余变量权重小于平均值,分类错误率最高为16.3%,最低为13.5%。测试样本的安全评价值达到优良,比层次分析法高12.1个百分点。研究结果表明,新建评价模型更加灵活机动,评价可信度更高,其应用实施能有效提高公共资源管理水平,为建设节约型校园奠定基础。

To build conservation-oriented campus as the base of scientific research and talent training, and reach the goal of people-orientation, resource saving, friendly environment and benign ecological circulation, and rapidly and accurately eveluate college public resources security operation to promote the reasonable management and utilization of resources, the similarities of random forest, AHP and grey correlation method were studied, and then the GRA-RF security evaluation model and evaluation index system were set up to classify and evaluate security variables. As the machine learning shown, the highest weight value was the medical treatment and antiepidemic, which was followed successively by fire protection measures, dining environment, accommodation, inspection and handling, learning and office, school gate management, institutions, and motor vehicles management. The weight values of the other variables were lower than the average. The highest classification error rate was 16. 3% and the lowest was 13. 5%. The evaluating values of sample were good, which were 12. 1% higher than that of AHP. The experiment results showed that the new evaluating model was more flexible and its reliability much higher, which can promote public resources management efficiently and lay the foundation for conservation-oriented college.