当前位置: X-MOL 学术J. Vis. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
A visual uncertainty analytics approach for weather forecast similarity measurement based on fuzzy clustering
Journal of Visualization ( IF 1.7 ) Pub Date : 2021-01-03 , DOI: 10.1007/s12650-020-00709-z
Renpei Huang , Li Chen , Xiaoru Yuan

Abstract

Forecast calibration methods based on historical similar atmospheric state are effective means weather forecast accuracy. Conventional approaches search similar forecasts on the basis of predefined similarity formulas and provide calibration recommendations to forecasters. However, these approaches ignore the uncertainty of similarity measurement, which affects calibration efficacy significantly. This study proposes a similarity weight adaptive algorithm for high-dimensional data on the basis of fuzzy clustering to characterize the uncertainty of similarity measurements. Without prior knowledge, the algorithm computes the uncertainty of the similarity between data in the fuzzy set space iteratively on the basis of membership and then determine weight distribution by maximizing the differentiating ability of each dimension. This study further presents a visual analysis framework on the basis of the weight adaptive algorithm for the exploration of uncertainty in meteorological data and the optimization of similarity measurement method. This framework has coordinated views and intuitive interactions to enable the visualization of the similarity uncertainty distribution and support the iterative visual analysis of similarity weight distribution in each dimension that combines domain knowledge. We illustrate a case study using real-world meteorological data to verify the efficacy of the proposed approach.

Graphic abstract



中文翻译:

一种基于模糊聚类的视觉不确定性分析天气预报相似度测量方法

摘要

基于历史相似大气状态的预报标定方法是有效的天气预报准确度。常规方法基于预定义的相似性公式搜索相似的预测,并向预测者提供校准建议。但是,这些方法忽略了相似性测量的不确定性,这极大地影响了校准功效。提出了一种基于模糊聚类的高维数据相似度加权自适应算法,以表征相似度测量的不确定性。在没有先验知识的情况下,该算法根据隶属关系迭代计算模糊集空间中数据之间相似性的不确定性,然后通过最大化各个维度的微分能力来确定权重分布。本研究进一步提出了一种基于权重自适应算法的可视化分析框架,用于探索气象数据的不确定性和相似性测量方法的优化。该框架具有协调的视图和直观的交互作用,以实现相似性不确定性分布的可视化,并支持对结合领域知识的每个维度中相似性权重分布的迭代视觉分析。我们举例说明了使用实际气象数据来验证所提出方法的有效性的案例研究。该框架具有协调的视图和直观的交互作用,以实现相似性不确定性分布的可视化,并支持对结合领域知识的每个维度中相似性权重分布的迭代视觉分析。我们举例说明了一个使用现实世界气象数据的案例研究,以验证所提出方法的有效性。该框架具有协调的视图和直观的交互作用,以实现相似性不确定性分布的可视化,并支持对结合领域知识的每个维度中相似性权重分布的迭代视觉分析。我们举例说明了一个使用现实世界气象数据的案例研究,以验证所提出方法的有效性。

图形摘要

更新日期:2021-01-03
down
wechat
bug