摘要 |
The dynamic determination of continuous split intervals for decision trees, without sorting, is disclosed. In one embodiment, the continuous split intervals are determined based on the mean of a relevant sub-set of the set of training data. In another embodiment, they are based on the mean and the standard deviation of the relevant sub-set. In another embodiment, the intervals are based on the median instead. In still another embodiment, they are based on a number of percentiles of the relevant sub-set of the set of training data.
|