Updatable Random Forests

Question

I use Random Forests for data forecasting and would like to continuously update the trained model using only new data instead of periodically re-training the model on the whole historical data. It seams to me that RF is not built to be updatable but ?

I use Weka's RF but any theoretical solution or experience on another solution could be of great help. Thank you

Answer 1

Weka's RF indeed doesn't have an option for dynamic retrain, but there are other algorithms (variations of the original RF) which have this option. This is one example: http://www.ymer.org/amir/software/online-random-forests/