机场噪声的关联规则挖掘及信念网络可视化表示研究

发布时间：2018-05-17 08:29

本文选题：机场噪声 + 关联规则　；参考：《中国民航大学》2015年硕士论文

【摘要】：机场噪声是多种噪声影响因素综合作用的结果。借助机场噪声历史数据集挖掘机场噪声与各噪声影响因素之间关系,有利于预测在各种噪声影响因素组合情况下噪声的可能影响程度,从而为制定合理的减噪措施提供科学依据。针对机场噪声与其影响因素之间无法简单地依靠建立函数模型来表达其复杂的非线性依赖关系的难题,本文在对机场噪声的每个影响因素进行定性和定量分析的基础上,挖掘机场历史噪声数据集中隐含的噪声与各种噪声影响因素之间的关联规则,并基于信念网络对其可视化表示。在对现有的关联规则兴趣度模型深入研究的基础上,本文提出通过兴趣规则前、后件的约束来避免挖掘无用规则的思想,并设计了相应算法BACIRC-FP-Growth-Cube(Based on Antecedent and Consequent of Interesting Rules Constrained FP-Growth-Cube),从而克服了目前大多数算法仅利用兴趣度过滤规则的弱点。该算法首先利用FP-Growth-Cube算法挖掘出频繁项集,再将兴趣规则前、后件与频繁项集相匹配,指导频繁项集定向生成用户感兴趣的关联规则,一定程度避免了关联规则作为无监督学习的盲目性。最后,利用机场噪声历史数据集的实验验证结果表明,与FP-Growth-Cube算法相比,BACIRC-FP-Growth-Cube算法更加高效,很大程度上减少甚至避免了无用规则的产生。为了综合表达机场噪声和其影响因素之间的关系,本文提出了基于信念网络的关联规则可视化表示方法,克服了关联规则无法表达不同规则之间联系的弱点。该方法通过拓扑结构学习和网络参数学习,将机场噪声和其影响因素图形化地表达为一个信念网络。关联规则本身表达数据中蕴含的知识,可为构建信念网络提供有效的先验知识,构建的网络具有较高的可接受度。最后,从机场降噪措施之一的机型选择角度,并与权威噪声预测软件INM相对比,表明了该表示方法的可行性和有效性。
[Abstract]:Airport noise is the result of the comprehensive effect of various noise factors. Using the historical data set of airport noise to excavate the relationship between the airport noise and the influence factors of the noise is helpful to predict the possible influence degree of the noise under the combination of various noise factors, so as to provide a scientific basis for the formulation of reasonable noise reduction measures. The field noise and its influencing factors can not simply rely on the establishment of a function model to express the complex nonlinear dependence. On the basis of qualitative and quantitative analysis of each influence factor of the airport noise, this paper excavates the correlation between the hidden noise of the airport historical noise data set and the influence factors of various noises. On the basis of the in-depth study of the existing interest model of the association rules, this paper proposes the idea of avoiding unuseld rules through the constraints of the rules of interest and the latter, and designs the corresponding algorithm BACIRC-FP-Growth-Cube (Based on Antecedent and Consequent of Interesti). Ng Rules Constrained FP-Growth-Cube), which overcomes the weakness that most algorithms use only interest filtering rules at present. Firstly, the algorithm uses FP-Growth-Cube algorithm to excavate frequent itemsets, and then matches the rules of interest, and then matches the frequent itemsets, directing the frequent itemsets to generate the association rules of interest of the users. Degree avoids the blindness of association rules as unsupervised learning. Finally, the experimental verification of the historical data set of airport noise shows that compared with the FP-Growth-Cube algorithm, the BACIRC-FP-Growth-Cube algorithm is more efficient and reduces the production of useless rules to a large extent. In order to express the airport noise and its influence, In this paper, a visual representation method of association rules based on belief networks is proposed, which overcomes the weakness of association rules that can not express the connections between different rules. This method expresses the noise of the airport and its influence factors into a belief network through topology learning and network parameter learning. The knowledge contained in the body expression data can provide effective prior knowledge for building the belief network, and the constructed network has high acceptability. Finally, from one of the noise reduction measures in the airport, the model selection angle is compared with the authoritative noise prediction software INM, which shows the feasibility and effectiveness of the method.
【学位授予单位】：中国民航大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：V351;TP311.13

【相似文献】