Transformers // Predicting next transaction based on sequence of previous transactions // Sequence2One task

Question

We are solving the following task. Our company has sequence of events like

DATA: 1000$ / Oranges / 11.00 am

500$ / Car wash / 03.00 pm

15$ / Flowers / 09.00 pm

TASK: The task is - To predict next transaction based on previous sequence of transactions

MY IDEA: I think generative models with similar to GPT architecture can perform well in this task. I want the model to consider (N) transactions given as an input as prompt and train model to output 3 categories separately (sum / category / time).

I was looking for code or approaches to solve similar tasks on the internet, but found nothing?

QUESTION:

1. Can anyone share of github code to solve a task like this?
2. Give suggestions on the approach and architecture?

Thx a lot :)

I think generative models with similar to GPT architecture can perform well in this task. I want the model to consider (N) transactions given as an input as prompt and train model to output 3 categories separately (sum / category / time).