How We Build A Model
The first step to building a model is to compile the right data. One set of data describes the individual and may contain age, gender, income, marital status, total debt, etc. The second set of data contains the examples of the behavior you are interested in modeling.
The combination of individual descriptive data and examples of desired behavior creates the development data set. Next, we clean, select, and transform the data to identify those elements worthy of further study.
Next, we consider and try several different statistical techniques to extract information from the data. Which technique is best depends on the data you have, the problem you are trying to solve and how the model is used. Our experience can help you navigate the pros and cons of each.
Our models are developed using a hold-out methodology. Models are built on one subset of the development data and validated on the other. This avoids over-fitting to the idiosyncrasies of the development data and increases the likelihood that the model will work well when applied to new data.