Time series decomposition

Question

I have a time series that I want to decompose. Dataset (train - dataframe) example (stock price):

        Date    Close
7389    2014-12-24  104.589996
7390    2014-12-26  105.059998
7391    2014-12-29  105.330002
7392    2014-12-30  105.360001
7393    2014-12-31  104.5700

Here is my code:

train_dec = copy.deepcopy(train)
train_dec.index = pd.to_datetime(train_dec['Date'])
train_dec.index.freq = 'D'

# Transform DataFrame into a Series
train_series = train_dec['Close']

train_decomposition = seasonal_decompose(train_series, model='additive')

train_trend = train_decomposition.trend
train_seasonal = train_decomposition.seasonal
train_residual = train_decomposition.resid

I tried without converting into Series and with it. Tried set up frequency to 'D'.

I keep getting errors such as:

ValueError: Inferred frequency None from passed values does not conform to passed frequency D

or

ValueError: You must specify a period or x must be a pandas object with a PeriodIndex or a DatetimeIndex with a freq not set to None

when I do not set frequency.

Maybe it is because the data have gaps (weekends) when there is no data point (stock price). Should I convert it to a weekly format? But how can I do this if there are gaps (e.g. if I have removed outliers)?

It must be something trivial but I can not see the solution.

Your help is greatly appreciated!

Answer 1

You need to specify the period when doing seasonal decomposition:

import pandas as pd
import numpy as np
from statsmodels.tsa.seasonal import seasonal_decompose
import matplotlib.pyplot as plt
import copy

data = {
    'Date': ['2014-12-24', '2014-12-26', '2014-12-29', '2014-12-30', '2014-12-31'],
    'Close': [104.589996, 105.059998, 105.330002, 105.360001, 104.5700]
}
train = pd.DataFrame(data)

train['Date'] = pd.to_datetime(train['Date'])
train.set_index('Date', inplace=True)

idx = pd.date_range(start=train.index.min(), end=train.index.max(), freq='D')
train = train.reindex(idx)

train['Close'] = train['Close'].ffill()

decomposition = seasonal_decompose(train['Close'], model='additive', period=3)  
fig = decomposition.plot()
plt.show()