Resolved: Frequency error on Time Series seasonal_decompose method

Question

Hello! When following the Time Series with Python course on the seasonality lecture, the method seasonal_decompose produces an error saying the frequency is not discernible.

'Index' object has no attribute 'inferred_freq'

I have tried the following troubleshooting methods:

Check to make sure the instruction was entered correctly
Reset the frequency of the series in the lecture from business days to just days to see if pd.infer_freq would pick it up
Used the original S&P data set column
Use integer index instead of date index

Any guidance would be much appreciated.

Answer 1

Hey Sia,

Thanks for reaching out!

This error occurs when the index column is not a series of dates with a set frequency (or when there are missing values). Hence, the easiest and mos straightforward solution is to fill out the data and set the frequency prior to setting the "date" column as the index.

df_comp = df_comp.asfreq('b')
df_comp = df_comp.fillna(method='ffill')

Alternatively, you can set the frequency of the dataset by setting a value for the "freq" argument of the method:

s_dec_training = sd(data_train.newspx, model = 'additive', freq = 'b')

However, that last part might not work on all versions of the statsmodels package, so I would stick with the former solution. Additionally, you can send us your Python Notebook, so we can see what's going on.

Best,
365 Vik

Resolved: Frequency error on Time Series seasonal_decompose method

Submit an answer