基于LOGISTIC和CART模型的风化影响因素研究

发布时间：2018-08-03 16:14

【摘要】：近几年来,随着旅游业的持续发展以及全球气候的变化,云冈石窟的依存环境发生了很大的变化。同时由于自然条件复杂且对云冈石窟风化作用的机理认识不足,云冈石窟石质文物的防风化保护一直是困扰我国文物保护工作的难题。因此有必要对以云冈石窟为例的石质文物风化影响因素进行科学的研究。本研究使用数据挖掘分类方法中的决策树CART分类算法模型和传统回归分析方法中的非条件Logistic回归模型,探讨影响石质文物风化的因素。并比较CART模型与Logistic模型之间结果的异同,进一步探讨CART模型与Logistic模型可以互为补充之处,以达到更准确的筛选风化影响因素的目的。首先,通过实地调研,专家咨询,文献研究等方法,了解石质文物风化机制,风化过程及可能的影响因素,并通过多种途径获取相关研究数据;其次,通过非条件Logostic回归分析,建立风化影响因素非条件Logistic模型,对模型结果进行分析验证;再次,在理清决策树CART算法原理的情况下,利用决策树CART算法建立风化影响因素决策树模型并对结果进行分析验证,并利用决策树模型提供的交互效应信息以及变量分类信息拟合新的Logistic回归模型,比较其与原Logistic模型筛选所得影响因素的差异;最后,利用各模型预测概率的ROC曲线下面积来评价模型分类的准确度,并说明CART决策树模型在筛选影响因素方面的优势以及改进后Logistic回归模型较改进前的准确性。本文将数据挖掘中的决策树分类方法应用到石质文物风化影响因素研究中,并与非条件Logistic模型相结合,能够更加充分的利用相关数据信息,得到更为全面的结果,这为石质文物风化影响因素研究提供了一种新的思路与方法,具有一定的理论和实际参考价值。
[Abstract]:In recent years, with the continuous development of tourism and the change of global climate, the dependent environment of Yungang Grottoes has changed greatly. At the same time, because of the complex natural conditions and lack of understanding of the mechanism of weathering in Yungang Grottoes, the weathering protection of Yungang Grottoes has been a difficult problem for the protection of cultural relics in China. Therefore, it is necessary to scientifically study the weathering factors of stone relics in Yungang Grottoes. In this study, the decision tree CART classification algorithm model and the unconditioned Logistic regression model in the traditional regression analysis were used to explore the factors affecting the weathering of stone relics. The similarities and differences of the results between CART model and Logistic model are compared, and the complementarities between CART model and Logistic model are discussed in order to screen the influence factors of weathering more accurately. First of all, through field investigation, expert consultation, literature research and other methods, to understand the weathering mechanism, weathering process and possible influencing factors of stone relics, and obtain the relevant research data through a variety of channels; secondly, through non-conditional Logostic regression analysis, The unconditioned Logistic model of weathering influencing factors is established, and the results of the model are analyzed and verified. Thirdly, in the case of clarifying the CART algorithm principle of decision tree, The decision tree model of weathering influencing factors is established by using decision tree CART algorithm and the results are analyzed and verified. The new Logistic regression model is fitted with the interactive effect information provided by decision tree model and the information of variable classification. Finally, the accuracy of model classification is evaluated by using the area under the ROC curve of the prediction probability of each model. The advantages of CART decision tree model in screening influencing factors and the accuracy of the improved Logistic regression model are also explained. In this paper, the decision tree classification method in data mining is applied to the study of the factors affecting the weathering of stone relics, and combined with the non-conditional Logistic model, it can make full use of the relevant data information and obtain more comprehensive results. This provides a new way of thinking and method for the study of the factors affecting the weathering of stone relics and has certain theoretical and practical reference value.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2010
【分类号】：K879.2;P512.1

【参考文献】