At Creme Global we have developed hundreds of successful models.
At the beginning of an analytics project we engage in workshops with our client to first establish what decisions need to be made. Once these decisions are established in principle we investigate what data is available to facilitate decision making. Finally we consider how these decisions can be better informed using data analytics.
We work with our client to gain access to all the necessary data. Then we undertake the process of cleaning the data. Our scientists at this point will identify any gaps and corner cases in the data, which may or may not require client collaboration to address. Finally we explore the data through visualisations to truly understand the data intuitively.
When our data scientists are comfortable with the data it is time to work on the model. We firstly identify inputs, outputs and key variables. We then draft the necessary number of candidate models in Python or R. Finally we carry out extensive experimentation, testing and validation to select the best model.
Finally we can use our model(s) to predict outcomes. We run our model to produce outputs. These outputs are analysed and visualised. For purely analytics-based projects these outputs often end up in a report (for software, see our process here). Our results are then stored and documented for reproducibility. We may then examine the potential for creating a production version of the model.