Optimus RP takes tree-based modeling a step further with patented algorithms that make it the fastest and most powerful recursive partitioning engine of its kind by employing true multi-way segmentation.  With multi-way splits, data is optimally segmented on all truly predictive, statistically significant cut points.  You are never forced to artificially create bins, nor are you ever limited by the deficiencies of simple binary splits.

While multi-way splits have been theorized for many years, they have historically been avoided due to inherent computational complexities; the number of k-way splits of n data points is O(n^k) – an exponentially huge search problem.
There are two algorithms in Optimus RP for finding the optimal splits. 

The first is based on an optimization of the original formulas which dramatically reduces computational complexity and time (in O(k n²) time). A second approximation (patent pending) algorithm solves the problem even faster, making it useable in an interactive mode on a standard PC. These algorithms also optimally deal with multi-way categorical splits and missing values.

Some claim that multi-way splits exhaust your data too soon, or that multiple binary splits will result in the same level of accuracy as true multi-way splits.  This is simply justifying the limitations of older tools and methods.  Settling for binary splits is no longer necessary – it is simply settling.  Consistently, and across a broad range of problems, optimal multi-way splits outperform binary splits in prediction performance.  Click here for an example.

Optimus RP also extends optimal multi-way segmentation to multivariate outcomes.