Budget Optimizer

[biggielutz]

Newer to Python and having trouble with the Budget Optimizer. I run the code:

Budget Optimization

prices = jnp.ones(mmm.n_media_channels)

n_time_periods = 10
budget = jnp.sum(media_data.mean(axis=0)) * n_time_periods
extra_features_forecast = extra_features_scaler.transform(extra_features_test)[:n_time_periods]

run budget optimization

solution = optimize_media.find_optimal_budgets(
n_time_periods=n_time_periods,
media_mix_model=mmm,
extra_features=extra_features_forecast,
budget=budget,
prices=prices,
media_scaler=media_scaler,
target_scaler=target_scaler)

budget, jnp.sum(solution.x * prices)

and get the error-
AttributeError Traceback (most recent call last)
Cell In[39], line 2
1 # both values should be almost equal
----> 2 budget, jnp.sum(solution.x * prices)

AttributeError: 'tuple' object has no attribute 'x'

any ideas on what I am doing wrong?

[thipokKub]

From the source code. The optimize_media.find_optimal_budgets return tuple of 3 items which are solution, kpi_without_optim, and starting_values. So I think your code should look like solution[0].x instead

[biggielutz]

Thanks! It hasn't worked for me yet- let me look at the source code again and try to get this right......

[biggielutz]

I used the demo https://github.com/google/lightweight_mmm/blob/main/examples/simple_end_to_end_demo.ipynb code for the optimizer and it worked perfect;y thatnks for your help....