What are the tuning parameters? How do I choose them?
EZPredict uses proprietary mathematical algorithms for finding correct predictions. It produces classifications from six different mathematical algorithms. Not all algorithms work equally well. It depends upon the nature of the data. For two of the six models, EZPredict provides users an ability to improve model predictions via tuning parameters. These tuning parameters take values between 0 and 1. We recommend that you try to improve the classifications from these two models by choosing values in halves, starting at .5, such as 0.5, 0.25, 0.125, etc. until you are satisfied. This will help them optimize classifications from models 1 and 2 quickly. It is quite possible that predictions from tuned models 1 and 2 could be better than predictions from the other models in EZPredict. In addition, EZPredict provides an overall tuning parameter for all the 6 models. It is typically set to 0. We strongly recommend that this parameter be changed only in small increments from 0, say 0.25%, .5%,
