![IconGems-Print](images/IconGems-Print.png)
199 – SPEED: Data Expo
Applying Functional Data Analysis and Clustering Methods on Weather Forecast Data in the U.S.
Yifan Wu
Simon Fraser University
Ying (Daisy) Yu
Simon Fraser University
Chuyuan (Cherlane) Lin
Simon Fraser University
The objective of this study is to investigate the potential covariates correlated to the weather prediction performance in the U.S, especially to explore the spatial and time effects in the prediction accuracy. We performed the functional principal component analysis (FPCA) and time series clustering techniques to divide 50 U.S. states into clusters. Cluster-specific characteristics of weather prediction performance were visually detected and cluster-to-cluster differences were quantified in order to identify this most and least predictable U.S. states. Then we conducted a functional analysis to capture the main pattern of variance in the prediction error over time and further investigate how other weather-related variables correlate with the prediction accuracy.