BonArea

We prototyped the 2 solutions, but we were focused on the second solution.

The first solution could have been carried through temporal series (ARIMA) but we have no expertise in this. We have been able to match temporal data with festivities in Catalunya and weekends.

For the second solution we prototyped a Neural Network using Dropout layers to eliminate correlations within data. Previous to the model, we have worked on data processing by calculating the total amount of money spent in the previous purchase of the client. We have done a data analysis, showing correlations and the training of the model. However, as the results were not good enough, we do not provide testing prediction values.

Throughout more work and time, we would have been able to deliver enough good testing results to deploy a model