With the help of artificial intelligence, Adtrac helps its customers to provide accurate forecasts of people flows in order to plan efficient campaigns.
General masking requirements during the Corona pandemic, as well as any software updates, can severely disrupt the recording of data. As a result, the forecasted numbers are no longer accurate.
To remedy this, it is necessary to adjust the model to the new situation as quickly as possible. This is difficult because we often observe annual seasonalities in the data, which can only be fed into the model by recording at least one cycle of data.
Adtrac has now found a way to solve this problem using a method called transfer learning. This involves taking a pre-trained model that already has the seasonalities and then adjusting it with the new data. As we have shown, this leads to the result that we can map the change in the data and still retain the seasonality.
For more details, see our blog post on Medium: