Is it possible to apply seasonal_decompose() after a groupby() where date is not the index of the data frame?

Did one of the already posed solutions work? If so or you found a different solution please share. I tried each without success, but I'm new to Python so probably missing something.

Here is what I came up with, using a for loop. For my dataset it took 8 minutes to decompose 20 million rows consisting of 6,000 different subsets. This works but I wish it were faster.

segments = set(frame['Segment ID'])
data = pd.DataFrame([])
for s in segments:
    df = frame[frame['Segment ID'] == s].set_index('Date Time').resample('H').mean()
    comp = sm.tsa.seasonal_decompose(x=df['Travel Time(Minutes)'], period=24*7, two_sided=False) 
    df = df.join(comp.trend).join(comp.seasonal).join(comp.resid)

    #optional columns with some statistics to find outliers and trend changes
    df['resid zscore'] = (df['resid'] - df['resid'].mean()).div(df['resid'].std())
    df['trend pct_change'] = df.trend.pct_change()
    df['trend pct_change zscore'] = (df['trend pct_change'] - df['trend pct_change'].mean()).div(df['trend pct_change'].std())

    data = data.append(df.dropna())

where you have lambda x: x.groupby(..., you don't have anything to group; you are telling it to group a row (I believe). You can try a setup like this, perhaps

Here you define a function to act on the group you are sending via the apply() method. Then you should be able to use your original code.

I have not tested this, but I use this setup quite often to work on groups.

def trend_function(x):
    # do your lambda function here as you are sending each grouping
      x.assign(
      trend = lambda x: x["value"].mask(~x["value"].isna(), other = 
      seasonal_decompose(x["value"][~x["value"].isna()], model='aditive', extrapolate_trend='freq').trend))
    return x

dfnew = df.groupby('time_series_id').apply(trend_function)

use extrapolate_trend='freq' as a parameter. you add the trend, seasonal, and residual to a dictionary and plot the dictionary from statsmodels.graphics import tsaplots import statsmodels.api as sm

date=['2015-08-01','2015-08-02','2015-08-03','2015-08-04','2015-08-01','2015-08-02','2015-08-03','2015-08-04']
time_series_id=[0,0,0,0,1,1,1,1]
value=[0,1,2,3,2,3,4,5]

df=pd.DataFrame({'date':date,'time_series_id':time_series_id,'value':value})

df['date']=pd.to_datetime(df['date'])
df=df.set_index('date')
print(df)

index_day = df.index.day
value_by_day = df.groupby(index_day)['value'].mean()
fig,ax = plt.subplots(figsize=(12,4))
value_by_day.plot(ax=ax)
plt.title('value by month')
plt.show()

df[['value']].boxplot()
plt.show()

fig,ax = plt.subplots(figsize=(12,4))
df[['value']].hist(ax=ax, bins=5)
plt.show()
fig,ax = plt.subplots(figsize=(12,4))
df[['value']].plot(kind='density', ax=ax)
plt.show()

plt.clf()
fig,ax = plt.subplots(figsize=(12,4))
plt.style.use('seaborn-pastel')
fig = tsaplots.plot_acf(df['value'], lags=4,ax=ax)
plt.show()

decomposition=sm.tsa.seasonal_decompose(x=df['value'],model='additive',     extrapolate_trend='freq', period=1)
decomposition.plot()
plt.show()

decomposition_trend=decomposition.trend
ax= decomposition_trend.plot(figsize=(14,2))
ax.set_xlabel('Date')
ax.set_ylabel('Trend of time series')
ax.set_title('Trend values of the time series')
plt.show()

I changed the first piece of code according to my scinario. Here's my code and attached output

data = pd.DataFrame([])

segments = set(subset['Planning_Material'])
for s in segments:
    df = subset[subset['Planning_Material'] == s].set_index('Cal_year_month').resample('M').sum()
    comp = sm.tsa.seasonal_decompose(df) 
    df = df.join(comp.trend).join(comp.seasonal).join(comp.resid)
    df['Planning_Material'] = s
    data = pd.concat([data,df])
data = data.reset_index()
data = data[['Planning_Material', 'Cal_year_month', 'Historical_demand', 'trend','seasonal','resid']]

data