all_data = pd.concat([xtrain , xtest]).drop(['time', 'code', 'first_prch'], axis=1).iloc[:1000]

target = all_data.groupby('id').mean()['target']

mean = all_data.drop('target' , axis=1).groupby('id').mean()
std = all_data.drop('target' , axis=1).groupby('id').std()
median = all_data.drop('target' , axis=1).groupby('id').median()
max_ = all_data.drop('target' , axis=1).groupby('id').max()
min_ = all_data.drop('target' , axis=1).groupby('id').min()
sum_ = all_data.drop('target' , axis=1).groupby('id').sum()
dif = max_ - min_
last_activity = all_data.sort_values('day_from_begin').groupby(['id']).apply(lambda xf: xf.iloc[-1])

all_data = pd.concat([target, std, mean , median , max_ , min_, sum_, dif, last_activity], axis=1)

cols=pd.Series(all_data.columns)
for dup in all_data.columns.get_duplicates(): cols[all_data.columns.get_loc(dup)]=[dup+'.'+str(d_idx) if d_idx!=0 else dup for d_idx in range(all_data.columns.get_loc(dup).sum())]
all_data.columns=cols

x_train = all_data[all_data['target'].notnull()]
x_test = all_data[all_data['target'].isnull()]