Sensitivity analysis of a (scikit-learn) machine learning model

sensitivity_analysis_example.py

from sklearn.datasets import make_regression
import pandas as pd
from xgboost import XGBRegressor
import matplotlib.pyplot as plt
import seaborn as sns

X, y = make_regression(n_samples=500, n_features=4, n_informative=2, noise=0.3)
X = pd.DataFrame(X, columns=['A', 'B', 'C', 'D'])

model = XGBRegressor()
model.fit(X, y)

class Simulate:
    def __init__(self, obs, var):
        self.obs = obs
        self.var = var

    def simulate_increase(self, model, percentage):
        baseline = model.predict(self.obs)
        plus = {}
        for ivar in self.var:
            X_plus = self.obs.copy()
            X_plus[ivar] = X_plus[ivar] + X_plus[ivar] * (percentage / 100)
            plus[ivar] = model.predict(X_plus)
        b = pd.DataFrame(
            plus, index=['simulated'
                         ]).T.reset_index().rename(columns={'index': 'test'})
        b['baseline'] = baseline[0]
        return b

    @staticmethod
    def plot_simulation(d, **kwargs):
        fig, ax = plt.subplots()
        sns.barplot(x='test', y='simulated', data=d, palette='deep', ax=ax)
        ax.axhline(d['baseline'].values[0], color='grey', linestyle='--', linewidth=2)
        ax.plot([0, 0], [-100, -100], color='grey', linestyle='--', linewidth=2, label='baseline')

        maxi = int(d['simulated'].max() + d['simulated'].max() * 0.1)
        mini = int(d['simulated'].min() - d['simulated'].min() * 0.1)
        ax.set_ylim([mini, maxi])

        ax.set_xlabel('Simulated variables')
        ax.set_ylabel('Target value')
        ax.set_title(kwargs.get('title'))
        ax.legend()

        ax.grid(axis='y', linewidth=.3)
        sns.despine(offset=10, trim=True)
        plt.tight_layout()
        plt.show()


VAR_OPTIMIZE = ['A', 'B', 'C']
PERC = 5
ROW = X.iloc[[29]]

S = Simulate(obs=ROW, var=VAR_OPTIMIZE)
d = S.simulate_increase(model=model, percentage=PERC)
S.plot_simulation(d, title=f'Impact of a {PERC}% increase of {VAR_OPTIMIZE} in target value')