pipeline.py

import numpy as np
import pandas as pd
from sklearn.decomposition import FastICA
from sklearn.ensemble import GradientBoostingClassifier
from sklearn.model_selection import train_test_split
from sklearn.pipeline import make_pipeline, make_union
from sklearn.tree import DecisionTreeClassifier
from tpot.builtins import StackingEstimator
from tpot.export_utils import set_param_recursive

tpot_data = pd.read_csv('cleaned_data.csv')
features = tpot_data.drop('Classification', axis=1)
training_features, testing_features, training_target, testing_target = \
            train_test_split(features, tpot_data['Classification'], random_state=42, test_size=0.2)

# Average CV (default=5) score on the training set was: 0.8188888888888888
exported_pipeline = make_pipeline(
    FastICA(tol=0.05),
    StackingEstimator(estimator=GradientBoostingClassifier(learning_rate=0.1, max_depth=7, max_features=0.3, min_samples_leaf=14, min_samples_split=16, n_estimators=100, subsample=0.25)),
    StackingEstimator(estimator=DecisionTreeClassifier(criterion="entropy", max_depth=8, min_samples_leaf=11, min_samples_split=12)),
    StackingEstimator(estimator=DecisionTreeClassifier(criterion="entropy", max_depth=9, min_samples_leaf=14, min_samples_split=5)),
    GradientBoostingClassifier(learning_rate=0.01, max_depth=5, max_features=0.55, min_samples_leaf=1, min_samples_split=13, n_estimators=100, subsample=0.5)
)
# Fix random state for all the steps in exported pipeline
set_param_recursive(exported_pipeline.steps, 'random_state', 42)

exported_pipeline.fit(training_features, training_target)
results = exported_pipeline.predict(testing_features)

# score
from sklearn.metrics import accuracy_score

print(accuracy_score(testing_target, results, normalize=True))

# export as pkl
import pickle

pickle.dump(exported_pipeline, open('model.pkl', 'wb'))