LILE, coming soon.
My latest project is a program to identify groupings in large amounts of data. A bit abstract, I admit, but it has wide-ranging applications. By identifying groupings, you free yourself from the requirement that all data be numeric. Your data models no longer need to be straight, or even continuous, lines.
Your predictive power goes way, way up.
At this stage of development it works, if a bit slowly, and preliminary results are promising. I am testing on a variety of datasets used for machine learning research kindly provided by the folks at Carnegie-Mellon (these are public datasets, I don’t have any big university connections right now 
). How promising? LILE is predicting bodyfat percentages ten percent better than multiple linear regression, and customer churn fifty percent better.
I am very excited, and hope to have a private alpha release sometime in the next three weeks. Contact me if you’d like to be part of it!
