1. Model Development
Select Modeling Technique
Generate Test Design
Build the Model
Assess the Model
Evaluate Results

The senior predictive model developer assigned to your project will be well experienced with a variety of modeling tools and will select the one(s) most appropriate for your circumstance.  Our approach requires that all models be developed using at least two different tools in order to ensure your model is as accurate as possible. 

 The selection of modeling technique can also significantly impact project level-of-effort and your ability to maintain and utilize the information into the future.  Wherever possible, we favor tools that require minimum preprocessing of data so that it is maintained in its native state to the extent possible and ‘intuitive preprocessing’ is eliminated as a source of weakness in the process.

 Additionally, all of our models are based on three sample data sets instead of the two that are more common.  All models require a training and testing data set.  The model is developed on the training set and tested against the testing set to minimize the possibility of over-fitting.  The PreMo approach always uses a third data set typically called a validation set or the ‘unseen’ data set.  Once the model is fully developed, its accuracy is evaluated against this validation data set which was not used in development.  This is an accurate reflection of the way the model will behave in the real world.

 We are very focused on model accuracy.  While most clients are happy to have a good model, at PreMo we believe we can make almost any existing model better.  Read about the PreMo Accuracy Challenge and why we believe we can make this claim.