changed question about minimum success rate
This commit is contained in:
@@ -1,5 +1,13 @@
|
||||
{
|
||||
"cells": [
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "74682f1a",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"# Done by Aviolat Charline, Bach Joachim and Marino Gabriel"
|
||||
]
|
||||
},
|
||||
{
|
||||
"cell_type": "markdown",
|
||||
"id": "ad0d40d6",
|
||||
@@ -106,7 +114,7 @@
|
||||
"id": "4bb03365",
|
||||
"metadata": {},
|
||||
"source": [
|
||||
"If we only take the priors, then A has 5/6 chances to be chosen vs B 1/6. The minimum accuracy is then 5/6, because the system will choose based on the 5/6 and 1/6, which will yeild 5/6 good results"
|
||||
"If we only take the priors, then the posterior probability only depends on it. The system will chose the highest posterior probability, so the highest prior because it is all it has. This means it will always choose the class A. If the repartition of the test set is the same as the training set, then always choosing A will give a 5/6 success rate, which will be all the correct A and all the missed B. If the test set is balanced, the success rate will be 50% because it will find all the A and miss all the B. Finally, if the system is unbalanced in the other way, the success rate will only be the portion of the A class in comparaison to the B class. The absolute minimum is then how low the portion of A can be compared to B in the test set."
|
||||
]
|
||||
},
|
||||
{
|
||||
|
||||
Reference in New Issue
Block a user