I have a dataset of time-series data with almost 100 stores. Time period is couple of years (but not for all stores - there are some older and newer ones).

To perform predictions of future sales (first approach with statsmodels) I decided to change that df to:

I am able to add those columns one by one but I fail with creating some kind of loop. My approach for doing it one by one:

store_23 = df[df['Store_num'] == 23].copy()
store_23.set_index(store_23['date'], inplace = True)
store_23.drop(['Store', 'date'], axis = 1, inplace = True)
store_23.columns = ['23_sales']

I was thinking also of creating a list of sales names (column names):

df_list = df['Store_num'].unique()
y_list = [str(num)   "_y" for num in df_list]

and than creating new df and adding data using groupby but I also failed in this approach.

Can anyone give me some suggestions ? Maybe my approach is completly wrong ?

CodePudding user response：

iiuc you want df.pivot:

df = pd.DataFrame(
    {
        'store_num':[0,0,0,1,1,1,2,2,2],
        'date':[1,2,3]*3,
        'sales':np.random.randint(0,10,9),
    }
)

df.pivot(index='date', columns='store_num')