CV: 27.6 LB: 29.33

cv 26 lb 29

CV: 30.26 LB: 32.35

you may use data from the same place in both train and test

try this:

from sklearn.cluster import KMeans

tr=train['Place_ID'].unique()[x:]

te=train['Place_ID'].unique()[:x]

tr=train[train['Place_ID'].isin(tr)]

te=train[train['Place_ID'].isin(te)]

X_train,y_train=tr.drop(columns=['Place_ID X Date', 'Date', 'Place_ID', 'target', 'target_min',

'target_max', 'target_variance', 'target_count','target_diff']),tr['target']

X_test,y_test=te.drop(columns=['Place_ID X Date', 'Date', 'Place_ID', 'target', 'target_min',

'target_max', 'target_variance', 'target_count','target_diff']),te['target']

CV: 24.52 LB: 30.26

Before giving your CV score, what is your validation strategy so that everyone can compare ?