PiML Toolbox for Model Development and Validation: Low-code Demo¶

PiML (Python Interpretable Machine Learning) is a new Python toolbox for IML model development and validation. Through low-code automation and high-code programming, PiML supports various machine learning models in the following two categories:

Inherently interpretable models:
1. EBM: Explainable Boosting Machine (Nori, et al. 2019; Lou, et al. 2013)
2. GAMI-Net: Generalized Additive Model with Structured Interactions (Yang, Zhang and Sudjianto, 2021)
3. ReLU-DNN: Deep ReLU Networks using the Aletheia Unwrapper (Sudjianto, et al. 2020)
Arbitrary black-box models，e.g.
1. LightGBM or XGBoost of varying depth
2. RandomForest of varying depth
3. Residual Deep Neural Networks

This example notebook demonstrates how to use PiML in its low-code mode for developing machine learning models, interpreting them and testing them. The toolbox has the following built-in datasets for demo purposes.

CoCircles classification data: simulated by sklearn.datasets.make_make_circles(n_samples=2000, noise=0.1); see details.
Friedman regression data: simulated by sklearn.datasets.make_friedman1(n_samples=2000, n_features=10, and noise=0.1); see details.
BikeSharing regression data from UCI repository: consisting of 17,389 samples of hourly counts of rental bikes in Capital bikeshare system; see details.
CaliforniaHousing regression data: consisting of 20,640 samples and 9 features, fetched by sklearn.datasets; see details. There are a raw version, a trim1 version (trimming only AveOccup) and a trim2 version (trimming AveRooms, AveBedrms, Population and AveOccup).
TaiwanCredit classification data fro UCI repository: consisting of 30,000 credit card clients in Taiwan from 200504 to 200509; see details. This data is subject to slight preprocessing.

Stage 0: Install PiML package on Google Colab¶

Run !pip install piml to install the latest version of PiML
In Colab, you'll need restart the runtime in order to use newly installed PiML version.

In [ ]:

!pip install PiML

Stage 1: Initialize an experiment, Load and Prepare data ¶

In [ ]:

from piml import Experiment
exp = Experiment(platform="colab")

In [ ]:

exp.data_loader()

In [ ]:

exp.data_summary()

In [ ]:

exp.data_prepare()

In [ ]:

exp.eda()

Stage 2. Train intepretable models ¶

In [ ]:

exp.model_train()

Stage 3. Explain and Interpret ¶

In [ ]:

exp.model_explain()

In [ ]:

exp.model_interpret()

Stage 4. Diagnose and compare¶

In [ ]:

exp.model_diagnose()

In [ ]:

exp.model_compare()

Stage 5. Register an arbitrary model ...¶

In [ ]:

# train_x, train_y, test_x, test_y, Xnames, yname = exp.get_processed_data() 

from lightgbm import LGBMRegressor
pipeline = exp.make_pipeline(LGBMRegressor(max_depth=7))
pipeline.fit() #train_x, train_y
exp.register(pipeline=pipeline, name='LGBM')

In [ ]:

exp.model_compare()