Hello, team! I'm working with S&P 500 index data that has datetime information in 5-minute intervals, like the following examples: "2024-04-24 093000-04:00, 2024-04-24 093500-04:00..." I tried using the freq='5min' parameter, but I'm getting an error. I think 'freq' param also have to contain 'b' (working days). How can I handle this?

Hello. I think your timestamp includes a timezone. Can you retry without including the timezone component? On my end, I was able to run the model on data with

freq='5min'

.

I tried without timezone and

freq

, but got this error

Exception: Could not infer frequency of ds column. This could be due to inconsistent intervals. Please check your data for missing, duplicated or irregular timestamps

On my end, I have a simulated dataset with irregular timestamps and a 5-min frequency (see image) and it still works. I used this code:

preds = nixtla_client.forecast(
    df=df,
    h=10,
    freq='5min',
    time_col='timestamp',
    target_col='price'
)

Make sure to have version 0.5.1 of

nixtla

installed. Then, the problem might be that you have duplicated timestamps.

it works, Thank you very much.