[python-package] monotonic constraints don't work

[yuanf8]

Description

I've trained a monotonic binary classifier with 'monotone_constraints_method': 'advanced', but as per the example below, it doesn't seem to work : (

Reproducible example

import lightgbm as lgb
import pandas as pd

bst = lgb.Booster(model_file='model_file_test.txt')
df = pd.read_csv('test_data.csv')
df['pred_score'] = bst.predict(df.drop(columns=['pred_score']))

In the test_data.csv, the two records only differ in feature columns col_13 and col 15.

	col_13	col_15
0	9	2
1	28.3156374	2.32535749

And in the model_file_test.txt,

feature_names=col_1 col_2 col_3 col_4 col_5 col_6 col_7 col_8 col_9 col_10 col_11 col_12 col_13 col_14 col_15 col_16 col_17 col_18 col_19 col_20 col_21 col_22 col_23 col_24 col_25 col_26 col_27 col_28 col_29 col_30 col_31
monotone_constraints=1 -1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 -1 -1 -1 -1 -1

Features col_13 and col_15 are constrained to be monotonic increasing.

However, the pred_scores from bst.predict() seem to decrease as the feature values increase.

	pred_score
0	0.41453699
1	0.41169457

Environment info

LightGBM version or commit hash: 4.0.0

Command(s) you used to install LightGBM:

python3.9 -m pip install lightgbm

[mayer79]

But the default method works, right?

[lorentzenchr]

@yuanf8 Could you provide a real minimal reproducible example, i.e. including the model fitting part which is crucial here? Otherwise, it’s impossible to hunt the error.