Frequently Asked Questions
Q4. What does a typical project look like?
There are a number of steps that our project with you will likely include. We’ve found that the acronym “L.E.T.M.E.” is useful in outlining these steps:
;)
L oad – Discovery Corps receives your data and loads it into our machines. Common formats (like DBF, MS Access, Excel, tab-delimited text, etc.) are preferred. We do some standard checks to insure that the data was exchanged successfully.
;)
E xplore – We will get to know your data by running preliminary distributions and getting other basic statistics. In this step we’ll also look for any data problems such as missing values, outliers, and anything else that needs a little cleanup.
;)
T ransform – Most projects involve creating new columns of data that are functions of your original columns. We often create binary (i.e., yes/no) and logarithmic transformations, as well as using our proprietary Discovery Corps Xmodeler [TM] technology to create “X transforms”. (To find out more about our Xmodeler technology, follow the link at the bottom of this page.)
;)
M odel –We use a variety of data mining/modeling techniques to capture hidden patterns in your data and to build models. Our mathematical mining and modeling tools learn from your data (that represents past experience) and create a mathematical formula or algorithm to predict what will happen in the future. This may take the form of a clustering model, a decision tree, a neural network, or some other model form that is best-suited to your needs.
;)
E valuate – Sometimes called “Validation”, at this stage we evaluate and test the performance of the predictive model we’ve created. After evaluation, if we feel we can create an even better model, we will loop back in the process armed with the insight we’ve gained thus far. When we’ve arrived at the optimum model we can build, we record the performance metrics so you’ll know the kind of success you can expect to see. This will be included in a project report and/or presentation to you.
For some clients, there is an additional Delivery/Deployment step:
;)
Delivery / Deployment – If the model is to run on a platform at your site, we will provide you with code that implements the scoring algorithm. We will be available to your team when they test the implementation to make sure it is giving correct results. This step also includes ad hoc support, performance monitoring, and software updates as appropriate. Alternatively, DCI provides scoring services so that you don’t have to do an implementation at your site. Call or email for more details.