from sklearn.model_selection import StratifiedKFold
skf = StratifiedKFold(n_splits=5,shuffle=True,random_state=777)

bin_ys = ys // 0.5

for idx, (trn_idx, val_idx) in enumerate(skf.split(xs, bin_ys)):
	print(f'now fold={idx:02d} split size is', skf.get_n_splits())
	trn_data = lgb.Dataset(xs.iloc[trn_idx], label=ys.values[trn_idx], categorical_feature=cats_index)
	val_data = lgb.Dataset(xs.iloc[val_idx], label=ys.values[val_idx], categorical_feature=cats_index)
	num_round = n_estimators
	clf = lgb.train(param, trn_data, num_round, valid_sets=[
		trn_data, val_data], verbose_eval=verbose, early_stopping_rounds=early_stopping_rounds)
	oof_predictions[val_idx] = clf.predict(xs.iloc[val_idx])
	test_predictions += clf.predict(XT) / skf.get_n_splits()
	....