논문 상세보기
pp.215-221
The purpose of this study was to incorporate Pakistan's climatic conditions into the road design process by performing a cluster analysis using collected climate data. Monthly time-series data for six climate variables—altitude, sea level, maximum temperature, minimum temperature, vapor pressure, and precipitation—were used to cluster 24 locations. Missing values were imputed using the Kalman filter, and hierarchical and k-medoid clustering analyses were performed based on the dynamic time warping (DTW) distance. By evaluating two to five clusters using six validity indices, the optimal number of clusters was determined to be two. the optimal two-cluster classification results were confirmed to be consistent between the two methods. When the clustering results were visualized on a map of Pakistan alongside the data, the clusters were divided into areas with relatively high and low altitudes. By classifying the regions of Pakistan into two clusters using time-series data of climate variables, this study highlights the distinct characteristics of each cluster. These findings suggest that management strategies tailored to the characteristics of each cluster can be applied to various fields.
1. 서론
2. 분석방법
2.1. 결측값 대체
2.2. DTW(Dynamic Time Warping)
2.3. 군집분석
3. 군집분석 결과
3.1. 데이터 개요
3.2. 계층적 군집분석
3.3. K-Medoids Clustering Analysis
3.4. 최적 군집분석
4. 결론
REFERENCES
