Performance Comparison of Multiple Similarity Algorithms in Meteorological Similarity Assessment

This paper conducts a detailed comparative analysis of the performance of nine commonly used similarity algorithms in meteorological similarity assessment. The study first constructs seven simple samples and calculates similarity values using both original data and 0-1 standardized data to evaluate the performance of each algorithm in terms of numerical and shape similarity. The results indicate that Hamming distance, Euclidean distance, root mean square distance, Li distance, and Yan criterion primarily reflect numerical similarity, while correlation coefficient, similarity coefficient, and Luo criterion place more emphasis on shape similarity. The Zhang criterion demonstrates a good balance between numerical and shape similarity. Data preprocessing has a minimal impact on the similarity rankings of most algorithms, but significantly affects the calculation reliability of the similarity coefficient and Yan criterion. This study provides technical references for the selection of similarity algorithms in meteorological similarity assessment.