Diferencia entre revisiones de «Índice de Czekanowski»
Ir a la navegación
Ir a la búsqueda
| Línea 6: | Línea 6: | ||
d = spp ausentes en común | d = spp ausentes en común | ||
| + | [http://people.revoledu.com/kardi/tutorial/Similarity/SimpleMatching.html Tutorial sobre medición de similitud] | ||
| + | [http://media.wiley.com/product_data/excerpt/61/04714696/0471469661.pdf Tutorial incompleto, buscar referencia. Explicación simple sobre índices de similitud y distancia] | ||
| + | <!-- | ||
| + | With binary variables, we traditionally focus on the notion of similarity rather | ||
| + | than distance (or dissimilarity). Consider two binary vectors x and y that consist | ||
| + | of two strings [xk], [yk] of binary data; compare them coordinatewise and do the | ||
| + | simple counting of occurrences: | ||
| + | number of occurrences when xk and yk are both equal to 1 | ||
| + | number of occurrences when xk = 0 and yk = 1 | ||
| + | number of occurrences when xk = 1 and yk = 0 | ||
| + | number of occurrences when xk and yk are both equal to 0 | ||
| + | |||
| + | These four numbers can be organized in a 2 by 2 co-occurrence matrix (contingency | ||
| + | table) that visualizes how “close” these two strings are to each other. | ||
| + | ::1 0 | ||
| + | ::1 a b | ||
| + | ::0 c d | ||
| + | Evidently the zero nondiagonal entries of this matrix point at the ideal matching | ||
| + | (the highest similarity). Based on these four entries, there are several commonly | ||
| + | encountered measure of similarity of binary vectors x and y. The simplest matching coefficient computes as the following ratio: | ||
| + | a + d | ||
| + | a + b + c + d | ||
| + | (1.4) | ||
| + | The Russell and Rao measure of similarity consists of the quotient | ||
| + | a | ||
| + | a + b + c + d | ||
| + | (1.5) | ||
| + | The Jacard index involves the case when both inputs assume values equal to 1: | ||
| + | a | ||
| + | a + b + c | ||
| + | (1.6) | ||
| + | The Czekanowski index is practically the same as the Jacard index, but by adding | ||
| + | the weight factor of 2, it emphasizes the coincidence of situations where entries | ||
| + | of x and y both assume values equal to 1: | ||
| + | 2a | ||
| + | 2a + b + c | ||
| + | (1.7) | ||
| + | --> | ||
[[Categoría:Glosario]] [[Categoría:Esbozo]] | [[Categoría:Glosario]] [[Categoría:Esbozo]] | ||
Revisión del 20:35 1 oct 2006
Cz =
a = ssp comunes b = exclusivas de grupo 1 c = exclusivas de grupo 2 d = spp ausentes en común
Tutorial sobre medición de similitud
Tutorial incompleto, buscar referencia. Explicación simple sobre índices de similitud y distancia
