from sklearn.model_selection import KFold
k_fold = KFold(n_splits=5, shuffle=True, random_state=2017)
pred_test_final = 0
5 folds are specified to show that 1 fold is for test and remaining is for train set. This is repeated for the next fold and so on. Minimum needs to be 2
for d_ind, v_ind in k_fold.split(train_x):
d_x, v_x = train_x.loc[d_ind, :], train_x.loc[v_ind, :]
d_y, v_y = train_y[d_ind], train_y[v_ind]
pred_test, model, evals_result = run_lgb(d_x, d_y, v_x, v_y, test_x) #Running LightGBM here
pred_test_final += pred_test
pred_test_final /= 5
pred_test_final = np.expm1(pred_test_final) #e(pred_test_final)-1