import lightgbm as lgb

np.random.seed(s)

d_train = lgb.Dataset(X_train, label = y_train)
d_submission = lgb.Dataset(X, label = y)

params = {}
params['learning_rate'] = 30
params['boosting_type'] = 'rf'
params['objective'] = 'multiclass'
params['metric'] = 'multi_logloss'
params['sub_feature'] = 1 # def 1
params['num_leaves'] = 350 # def 31
params['max_depth'] = -1 # -1
params['num_class'] = 20
params['min_data_in_leaf'] = 0 # def 20
# params['lambda_l2'] = 0.4


params['bagging_fraction'] = 0.9
params['bagging_freq'] = 10
params['feature_fraction'] = 0.9


clf = lgb.train(params, d_train, 100)

y_pred = clf.predict(X_train)
y_pred = [np.argmax(y_pred[i]) for i in range(len(y_pred))]

print(accuracy_score(y_train, y_pred))

y_pred = clf.predict(X_test)
y_pred = [np.argmax(y_pred[i]) for i in range(len(y_pred))]

print(accuracy_score(y_test, y_pred))