| Summary: | 碩士 === 國立嘉義大學 === 土木與水資源工程學系研究所 === 97 === In recent years, self-organizing map(SOM) is often used for cluster analysis, because it can project high-dimensional input space on a low dimensional topology, and preserve the original topology information and the inherent statistics characteristics. Principal component analysis (PCA) is a linear transformation technique that provides a smaller set of uncorrelated variables (called components) from a set of correlated variables while maintaining most of the information in the original data set. In this paper, a model based on the combination of PCA and SOM is applied to identify the homogeneous regions for regional frequency analysis. First, the annual maximum daily rainfall records from 127 gauges in Taiwan are available. Then PCA is applied to obtain the principal components. It is found that the first nine principal components explain over 80% of the information. Based on the transformed data resulting from PCA and the geographic characters of the gauges, the SOM is used to group the rain gauges into specific clusters. The 127 rain gauges are grouped into 17 clusters. The discordancy test and the heterogeneity test indicate that the 17 regions are sufficiently homogeneous. In addition, the results show that the SOM can identify the homogeneous regions more accurately as compared to the K-means method and Ward’s method. Finally, the SOM based on the original data is used to group the rain gauges. The results show that the SOM cannot identify the homogeneous regions more accurately. Therefore, the model based on the combination of PCA and SOM is recommended as an alternative to the identification of homogeneous regions for regional frequency analysis.
|
|---|
