a***a
发帖数: 434 | 1 比如默认是Inf 1.0, 0.9, 0.8, ...
怎么改为Inf, 1.0, 0.99, 0.98, 0.97, ...
求教~~thx！！ | D******n
发帖数: 2836 | 2 roc needs cutoff?
isnt roc designed to avoid using a cutoff?
【在 a***a 的大作中提到】
: 比如默认是Inf 1.0, 0.9, 0.8, ...
: 怎么改为Inf, 1.0, 0.99, 0.98, 0.97, ...
: 求教~~thx！！
| r*****a
发帖数: 27155 | 3 新手。。不太明白
FPR不是由cutoff定的么？
【在 D******n 的大作中提到】
: roc needs cutoff?
: isnt roc designed to avoid using a cutoff?
| D******n
发帖数: 2836 | 4 if u want FPR at a certain cutoff, it is FPR at a certain cutoff.
ROC is used to get the big picture of the performance of the algorithm
as the cutoff varies.
And AUC is used to describe the discriminatory power of the alogorithm.
【在 r*****a 的大作中提到】
: 新手。。不太明白
: FPR不是由cutoff定的么？
| r*****a
发帖数: 27155 | 5 我的意思是，如何修改x坐标的精度呢？那个plot应该是取不同FPR值来plot的吧？
比如FPR <- c(0.0, 0.1, 0.2, ..., 1.0)
可不可以指定classification的cutoff c？
就是这个：
probability > c, label <- 1
probability <= c, label <- 0
thx a lot~~
【在 D******n 的大作中提到】
: if u want FPR at a certain cutoff, it is FPR at a certain cutoff.
: ROC is used to get the big picture of the performance of the algorithm
: as the cutoff varies.
: And AUC is used to describe the discriminatory power of the alogorithm.
| D******n
发帖数: 2836 | 6 ROC curve is usually a zigzag plot. and the usual input is ur prediction and
the real pos/negative value. Such as
1.2 0
3.4 0
5.3 1
7.8 1
9.1 0
...
...
0 for negative 1 for positive. u can see u dont need to vary a cutoff to
plot it.
For example any value between 1.2 and 3.4 will be the same.
【在 r*****a 的大作中提到】
: 我的意思是，如何修改x坐标的精度呢？那个plot应该是取不同FPR值来plot的吧？
: 比如FPR <- c(0.0, 0.1, 0.2, ..., 1.0)
: 可不可以指定classification的cutoff c？
: 就是这个：
: probability > c, label <- 1
: probability <= c, label <- 0
: thx a lot~~
| r*****a
发帖数: 27155 | 7 哦！明白你意思了
thx a lot~~
and
【在 D******n 的大作中提到】
: ROC curve is usually a zigzag plot. and the usual input is ur prediction and
: the real pos/negative value. Such as
: 1.2 0
: 3.4 0
: 5.3 1
: 7.8 1
: 9.1 0
: ...
: ...
: 0 for negative 1 for positive. u can see u dont need to vary a cutoff to
|
|
