I am trying to predict customer spending for an X year period after t0. I train an svm model with transactions occurring before and on t0, on the cumulative spending of the customers after t0. I then use the model to make predictions on a test data set to assess the models accuracy. The test dataset is from 1 period (in the example below 1 day) in the future. How does an svm deal with newly acquired customers?
Question elaboration:
My training data is structured as follows:
>DT1
Id t.-3 t.-2 t.-1 t.0 Target.spending
1: 1 29 25 14 25 100
2: 2 NA 30 0 0 0
3: 3 NA 16 0 13 0
4: 4 NA NA 62 18 5
Customer 1 was acquired on t.-3 when he purchased worth 29$. His second purchase occurred on t.-2. etc... Customer 2 was acquired on t.-2 and only purchased on that day. Target.spending is the cumulative spending that occurred in the next four days (i.e. t.1+t.2+t.3+t.4).
Then, I predict the spending of the next four days using an svm:
train.model <- svm(Target.spending ~ . , data = DT)
Using a test data set from one day in the future (where t.0 (DT1) = t.-1 (DT2)), I will predict the target.spending (DT2) and assess the model accuracy. In the test data, on t.0 (DT2), customer 5 was acquired.
>DT2
Id t.-3 t.-2 t.-1 t.0 Target.spending
1: 1 25 14 25 10 103
2: 2 30 0 0 0 0
3: 3 16 0 13 0 0
4: 4 NA 62 18 4 1
5: 5 NA NA NA 9 30
How does the svm deal with new customer acquisitions?
