基于平行坐标和数据挖掘的P2P借贷金融数据可视化研究

发布时间：2019-01-27 18:11

【摘要】：本论文以真实的网络P2P借贷金融数据作为基础,结合计算美学中的可视化手段,与数据挖掘领域交叉结合,运用多种算法和表现形式,挖掘数据中的规律并直观展示影响借贷成功与否的影响因素。本研究以平行坐标等可视化手段作为呈现工具,对表现方法、计算手段进行了深入研究和改进,提出坐标轴上数据密度的概念,在保证效率的前提下对于绘制方式提出了如何呈现数据趋势的改进方法,加入多种交互手段以达到结果的灵活性和可变性,更结合使用动态画法产生更加具有交互效果的展示方式。为了呈现更加具有统计意义的结果,本研究还结合使用了Kmeans数据挖掘算法,进行数据聚类,并结合启发式算法对于初始数据和聚类参数进行改进,使得算法更加高效。对于挖掘出的结果,本研究将其灵活的融入平行坐标的交互展示中,使得数据的呈现方式更加多元且高效。
[Abstract]:This paper is based on the real network P2P lending financial data, combined with the visual means of computing aesthetics, and combined with the field of data mining, using a variety of algorithms and forms of representation. Mining the rules in the data and visualizing the factors that affect the success of the loan. In this study, parallel coordinates and other visualization tools are used as presentation tools. The representation method and calculation method are deeply studied and improved, and the concept of data density on coordinate axis is put forward. On the premise of ensuring efficiency, this paper puts forward an improved method on how to present the trend of data, and adds a variety of interactive methods to achieve the flexibility and variability of the results. Combined with the use of dynamic painting to produce a more interactive way of display. In order to present more statistically significant results, Kmeans data mining algorithm is used to cluster data, and heuristic algorithm is used to improve the initial data and clustering parameters to make the algorithm more efficient. For the results of mining, this study flexibly integrates them into the interactive display of parallel coordinates, which makes the presentation of data more diverse and efficient.
【学位授予单位】：天津大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP311.13;F724.6;F832.4

【相似文献】