Create tables using OptBinning with custom bins

[tomasleon2]

I want to use the library OptBinning to create tables with all the metrics, but under the assumption that I already have all the bins. I don't want to optimize the binning process, just want the tables with my current bins. Despite I've found a "solution" not sure if there's bug or something I'm missing in the parameters. Here my example: First I create a fake dataset:

import random
import pandas as pd
from datetime import datetime, timedelta
import pandas as pd
from optbinning import BinningProcess, OptimalBinning

Set seed for reproducibility

random.seed(42)

Generate fake data

data = {
'GB': [random.choice([0, 1]) for _ in range(2000)],
'Period': [(datetime(2021, 1, 1) + timedelta(days=random.randint(0, 731))).strftime("%m/%Y") for _ in range(2000)],
'Age': [random.randint(18, 80) if random.random() > 0.2 else None for _ in range(2000)],
'L6ag': [random.randint(0, 9) if random.random() > 0.2 else None for _ in range(2000)],
'L_3M': [chr(random.randint(65, 90)) if random.random() > 0.2 else None for _ in range(2000)],
'M36m': [random.randint(0, 1000) for _ in range(2000)],
'Balance': [random.randint(0, 100000) for _ in range(2000)]
}

Create DataFrame

df = pd.DataFrame(data)
df

Then, I want to for example create a table for Age using the following bins: custom_bins = [28, 37, 63, 67]
So, i use the following code:

Define your custom bins

custom_bins = [28, 37, 63, 67]

Define the binning object

optb = OptimalBinning(name="Age", dtype="numerical", user_splits=custom_bins)

Fit the binning object

optb.fit(df["Age"], df["GB"]) # GB is your target variable

optb.binning_table.build()

And I get the following table which miss the first bin (-inf to 28):
https://i.stack.imgur.com/TBfNt.png

If I try using the user_splits_fixed parameter to "force" each value on the bins, the result is even worse

Define your custom bins

custom_bins = [28, 37, 63, 67]
user_splits_fixed = [True, True, True, True]

Define the binning object

optb = OptimalBinning(name="Age", dtype="numerical", user_splits=custom_bins, user_splits_fixed=user_splits_fixed)

Fit the binning object

optb.fit(df["Age"], df["GB"]) # GB is your target variable

optb.binning_table.build()
https://i.stack.imgur.com/ZpZqD.png

Any help, would be more than appreciated

[tomasleon2]

Got the answer. monotonic_trend should be set to None, so there's no further calculations. It just create the table following the provided bins :)