Visibility and Meteorological Parameter Model Based on Rashomon Regression Analysis

Abstract

Atmospheric visibility is one of the critical indicators for meteorological characterization and environmental quality evaluation. This paper studies the influence of different meteorological parameters on atmospheric visibility, including seven main factors: temperature, humidity, wind speed, and atmospheric pressure. To establish a regression model of visibility calculation under the influence of multiple factors, this paper proposes a method named Rashomon principal component optimization regression. This paper specifically introduces the modeling and implementation of this method. The key is to solve the Rashomon coefficient, the uncertainty influence coefficient, and the regression dimension coefficient. This method employs principal component analysis to establish a loop algorithm that effectively selects different feature spaces. The main purpose is to reflect the multi-scale characteristics of the sample data, and not only consider the overall or local characteristics to deviate from the actual situation. In addition, the interaction between different factors is considered, and the analytic network process (ANP) model is used to reflect the uncertainty in the modeling. The proposed method benefits the future analysis and prediction of visibility based on meteorological data. Meanwhile, it provides theoretical support for big data problems under multiple factors.