Weird MAPE values about lightweight_mmm HOT 7 CLOSED

Hello @virithavanama !

If you are using spend as your media unit there is no need to include impressions. You can use either one but there is generally no benefit in having both.

With regards to the MAPE, I see some values are very close to 0 in your target, that can cause MAPE to give some very high values. It is one of the issues of MAPE.

from lightweight_mmm.

Hi Pablo,
Thank you so much for your response.

1. plot_media_baseline_contribution_area_plot is throwing an error saying it has both positive and negative values
2. When trying to run find_optimal_budgets for one year with last year's budget, it's saying the given budget is more than optimized.
   Thanks in advance

from lightweight_mmm.

Hello @virithavanama !

1. Can you show here the stacktrace and the error you are getting?
2. One year might be too long term for optimization (not that it cant be done, but results might need a bit more carefulness). Can you show the pre/post optimization plot?

from lightweight_mmm.

1. ValueError: When stacked is True, each column must be either all positive or all negative. Column 'PublisherDirect_x contribution' contains both positive and negative values
2. And is there any way to access the csv files for the plots?

from lightweight_mmm.

@yukaabe Could you take a look at this error? @virithavanama could you provide a colab or similar that reproduces the error? It can be with mock data.

For this one yes, you can get the dataframe for it.

from lightweight_mmm.

Hi Pablo,
Thank you so much for your response.
Sorry, I cannot provide the notebook because we are working with client data.
Regarding the CSV/data frame, I meant for all the plots not just the baseline plot

from lightweight_mmm.

No, at the moment all plots do not give the output data by default. Although it is something that we have considered.

from lightweight_mmm.