plot_period_transactions failed for BetaGeoBetaBinomFitter

[hmikelee]

Dear Sir,

While testing BG/BB models, I came across the following error when calling plot_period_transactions to verify the fitting results

TypeError: unsupported operand type(s) for +: 'int' and 'list'

Traceback below:

TypeError                                 Traceback (most recent call last)
<ipython-input-30-4d245b7ff070> in <module>()
      2 bbf.fit(summary_monthly['frequency'], summary_monthly['recency'], summary_monthly['n_periods'])
      3 print(bbf)
----> 4 plot_period_transactions(bbf)
      5 
      6 for period in summary_monthly['n_periods'].unique():

~/anaconda3/lib/python3.6/site-packages/lifetimes/plotting.py in plot_period_transactions(model, max_frequency, title, xlabel, ylabel, **kwargs)
     59 
     60     n = model.data.shape[0]
---> 61     simulated_data = model.generate_new_data(size=n)
     62 
     63     model_counts = pd.DataFrame(model.data["frequency"].value_counts().sort_index().iloc[:max_frequency])

~/anaconda3/lib/python3.6/site-packages/lifetimes/fitters/beta_geo_beta_binom_fitter.py in <lambda>(size)
    167         self.generate_new_data = lambda size=1: beta_geometric_beta_binom_model(
    168             # Making a large array replicating n by n_custs having n.
--> 169             np.array(sum([n_] * n_cust for (n_, n_cust) in zip(n_periods, weights))),
    170             *self._unload_params("alpha", "beta", "gamma", "delta"),
    171             size=size

TypeError: unsupported operand type(s) for +: 'int' and 'list'

BetaGeoBetaBinomFitter.fit went ok

I am attaching the screenshot of the notebook below

[SUBHAMILA]

code with error : self.generate_new_data = lambda size=1: beta_geometric_beta_binom_model( # Making a large array replicating n by n_custs having n. np.array(sum([n_] * n_cust for (n_, n_cust) in zip(n_periods, weights))), *self._unload_params("alpha", "beta", "gamma", "delta"), size=size )
i think the following changes may solve your problem .
try doing this :
you get that error because one is integer value and other is list.

try this :

self.generate_new_data = lambda size=1: beta_geometric_beta_binom_model( # Making a large array replicating n by n_custs having n. np.array(sum([n_] * sum(n_cust) for (n_, n_cust) in zip(n_periods, weights))), *self._unload_params("alpha", "beta", "gamma", "delta"), size=size )

[psygo]

@SUBHAMILA, please use three backticks (`) for code blocks, otherwise that is illegible.

[Trollgeir]

code with error : self.generate_new_data = lambda size=1: beta_geometric_beta_binom_model( # Making a large array replicating n by n_custs having n. np.array(sum([n_] * n_cust for (n_, n_cust) in zip(n_periods, weights))), *self._unload_params("alpha", "beta", "gamma", "delta"), size=size )
i think the following changes may solve your problem .
try doing this :
you get that error because one is integer value and other is list.

try this :

self.generate_new_data = lambda size=1: beta_geometric_beta_binom_model( # Making a large array replicating n by n_custs having n. np.array(sum([n_] * sum(n_cust) for (n_, n_cust) in zip(n_periods, weights))), *self._unload_params("alpha", "beta", "gamma", "delta"), size=size )

This did not work.

This did though:

The solution that worked was to modify row 169 in beta_geo_beta_binom_fitter.py from np.array(sum([n_] * n_cust for (n_, n_cust) in zip(n_periods, weights))) to np.array(sum([[n_] * >n_cust for (n_, n_cust) in zip(n_periods, weights)], []))

Credits to @cereusperuv (#259 (comment))