In allusion to the previous various forecasting verification results of each similarity measurement, starting from formulas, the fact that there existed the same features between analog deviation and Hamming distance ...In allusion to the previous various forecasting verification results of each similarity measurement, starting from formulas, the fact that there existed the same features between analog deviation and Hamming distance was analyzed and then proved. The limitations of analog deviation used in similarity forecast were discussed. The data from 96 different stations at 850 hPa in the East Asia during May 1 -30, 2010 were utilized to make similarity selection experiments among several frequently used similarity measurements and the new one proposed by the authors of this essay. The result indicated that analog deviation and Hamming distance were very much similar, with over 80% selected samples being the same. Analog deviation had the biggest difference with similarity coefficient, with over 70% selected samples being different. The new analog quantity was more similar to the related coefficient, with 60% selected samples being the same. Analog deviation and Hamming distance reflected how much the distances of the samples were similar to each other, while related coefficient and the new analog quantity reflected how much the shapes of the selected samples were similar to each other.展开更多
文摘In allusion to the previous various forecasting verification results of each similarity measurement, starting from formulas, the fact that there existed the same features between analog deviation and Hamming distance was analyzed and then proved. The limitations of analog deviation used in similarity forecast were discussed. The data from 96 different stations at 850 hPa in the East Asia during May 1 -30, 2010 were utilized to make similarity selection experiments among several frequently used similarity measurements and the new one proposed by the authors of this essay. The result indicated that analog deviation and Hamming distance were very much similar, with over 80% selected samples being the same. Analog deviation had the biggest difference with similarity coefficient, with over 70% selected samples being different. The new analog quantity was more similar to the related coefficient, with 60% selected samples being the same. Analog deviation and Hamming distance reflected how much the distances of the samples were similar to each other, while related coefficient and the new analog quantity reflected how much the shapes of the selected samples were similar to each other.
