Jackknife variance estimates for random forest

From HandWiki

In statistics, jackknife variance estimates for random forest are a way to estimate the variance in random forest models, in order to eliminate the bootstrap effects.

Jackknife variance estimates

The sampling variance of bagged learners is:

V(x)=Var[θ^(x)]

Jackknife estimates can be considered to eliminate the bootstrap effects. The jackknife variance estimator is defined as:[1]

V^j=n1ni=1n(θ^(i)θ)2

In some classification problems, when random forest is used to fit models, jackknife estimated variance is defined as:

V^j=n1ni=1n(t(i)(x)t(x))2

Here, tdenotes a decision tree after training, t(i) denotes the result based on samples without ith observation.

Examples

E-mail spam problem is a common classification problem, in this problem, 57 features are used to classify spam e-mail and non-spam e-mail. Applying IJ-U variance formula to evaluate the accuracy of models with m=15,19 and 57. The results shows in paper( Confidence Intervals for Random Forests: The jackknife and the Infinitesimal Jackknife ) that m = 57 random forest appears to be quite unstable, while predictions made by m=5 random forest appear to be quite stable, this results is corresponding to the evaluation made by error percentage, in which the accuracy of model with m=5 is high and m=57 is low.

Here, accuracy is measured by error rate, which is defined as:

ErrorRate=1Ni=1Nj=1Myij,

Here N is also the number of samples, M is the number of classes, yij is the indicator function which equals 1 when ith observation is in class j, equals 0 when in other classes. No probability is considered here. There is another method which is similar to error rate to measure accuracy:

logloss=1Ni=1Nj=1Myijlog(pij)

Here N is the number of samples, M is the number of classes, yij is the indicator function which equals 1 when ith observation is in class j, equals 0 when in other classes. pij is the predicted probability of ith observation in class j.This method is used in Kaggle[2] These two methods are very similar.

Modification for bias

When using Monte Carlo MSEs for estimating VIJ and VJ, a problem about the Monte Carlo bias should be considered, especially when n is large, the bias is getting large:

E[V^IJB]V^IJnb=1B(tbt¯)2B

To eliminate this influence, bias-corrected modifications are suggested:

V^IJUB=V^IJBnb=1B(tbt¯)2B
V^JUB=V^JB(e1)nb=1B(tbt¯)2B

References