TimeSeries Decomposition in Python with statsmodels and Pandas

TimeSeries-Decomposition.ipynb

I follow the steps that you follow and I got

In my case I have a huge amount of data so is difficult review this data
What do you suggest me?
How can I create more big graph?
Is there a procedure for big time series??

@Cyberguille.

res = sm.tsa.seasonal_decompose(centrumGalerie.Belegung.interpolate(),
                                freq=decompfreq,
                                model='additive')

You can plot a bigger graph by plotting each graph separately. For example,

import matplotlib.pyplot as plt 

fig, (ax1,ax2,ax3) = plt.subplots(3,1, figsize=(15,8))
res.trend.plot(ax=ax1)
res.resid.plot(ax=ax2)
res.seasonal.plot(ax=ax3)

Knowing the trend, seasonality and residuals, what do we do with them?

We want a trend that is fairly stable (like a straight line) before we do ARMA or ARIMA on the data, correct? We would do things like difference or log difference to address this issue?

What would we like from seasonal? Would we like it to be flat as well? How do we interpret the sinusoidal shape? And what cn we do to address seasonality?

@IanQS That's what I'm trying to figure out as well. I got the nice plots from seasonal_decompose, now have no idea how to proceed...

@Cyberguille

Just use that before the plot and you will be fine:

pylab.rcParams['figure.figsize'] = (14, 9)

when we get the decomposition components, how to predict the future steps?

Thanks for your comments,

I also want to know how can i use this data in ARIMA or FOURIER.

@Cyberguille

Just use that before the plot and you will be fine:

pylab.rcParams['figure.figsize'] = (14, 9)

Worked for me, thanks!

Hi..am trying to use your method in my project and am using many issues. Can anyone help me decompose my time series??

Just use that before the plot and you will be fine:

pylab.rcParams['figure.figsize'] = (14, 9)

Works perfectly!

@pratikask if you still need a help I can help you