Background: I am currently working on a class project in which I have to forecast weekly demand for a retailer. The time series is very minimal trend (most likely additive), and there is no seasonality. However, I am sure the data is cyclic. I tried to fit SARIMAX, and as expected the predictions were horrible. Next, I tried simple AR, and that gave somewhat reasonable forecast. I read a research paper which talks about using a variant of ARMA model for cyclic data.
Question / Request: Does anyone here knows about this specific application? I would be grateful if someone could share a python / R code to implement ARMA model for cyclic data.