"/X_train.p", "rb"))
y_train = pickle.load(
open(
"/home/felix/phd/feature_constraints/" + str(which_experiment) +
"/y_train.p", "rb"))
all_features = CandidateFeature(IdentityTransformation(-1),
numeric_representations)
all_standardized = CandidateFeature(MinMaxScalingTransformation(),
[all_features])
foreigner = np.array(X_train[:, 7])
gender = np.array(
['female' in personal_status for personal_status in X_train[:, 15]])
my_runner = Runner(c=1.0, sensitive=gender, labels=['bad', 'good'])
#my_runner = Runner(c=1.0, sensitive=foreigner, labels=['bad', 'good'])
model = xgb.XGBClassifier(objective="binary:logistic",
n_estimators=1000,
random_state=42)
history = []
my_pipeline = Pipeline([('f', all_standardized.pipeline), ('c', model)])
my_pipeline.fit(X_train, y_train)
gain_dict = my_pipeline.named_steps['c'].get_booster().get_score(
importance_type='gain')
#print(X_train[:,7])
#todo: measure TP for each group and add objective
#todo: try misclassification constraint
#todo: restart search
bit2results = {}
'''
foreigner = np.array(X_train[:,7])
gender = np.array(['female' in personal_status for personal_status in X_train[:,15]])
my_runner = Runner(c=1.0, sensitive=gender, labels=['bad', 'good'])
#my_runner = Runner(c=1.0, sensitive=foreigner, labels=['bad', 'good'])
'''
my_runner = Runner(c=1.0, experiment=which_experiment)
history = []
# define an objective function
def objective(features):
#_, test, pred_test, std_score, proba_pred_test = my_runner.run_pipeline(features, runs=1)
results = my_runner.run_pipeline(features, runs=1)
#print(features)
'''
assert len(y_test) == len(pred_test)
which_observation_should_be_predicted_correctly = 333#131
print(proba_pred_test[which_observation_should_be_predicted_correctly])
open(
"/home/felix/phd/feature_constraints/" + str(which_experiment) +
"/y_train.p", "rb"))
all_features = CandidateFeature(IdentityTransformation(-1),
numeric_representations)
all_standardized = CandidateFeature(MinMaxScalingTransformation(),
[all_features])
#foreigner = np.array(X_train[:,7])
#gender = np.array(['female' in personal_status for personal_status in X_train[:,15]])
#my_runner = Runner(c=1.0, sensitive=gender, labels=['bad', 'good'], experiment=which_experiment)
#my_runner = Runner(c=1.0, sensitive=foreigner, labels=['bad', 'good'])
my_runner = Runner(c=1.0, experiment=which_experiment)
#todo: put feature selection within pipeline!!!!
history = []
for c in [0.001, 0.01, 0.1, 1, 10, 100, 1000]:
model = LogisticRegression(penalty='l1',
C=1,
solver='saga',
class_weight='balanced',
max_iter=10000,
multi_class='auto')
my_pipeline = Pipeline([('f', all_standardized.pipeline), ('c', model)])
from fastsklearnfeature.interactiveAutoML.CreditWrapper import run_pipeline
import numpy as np
import pickle
from hyperopt import fmin, tpe, hp, STATUS_OK, Trials
from hyperopt import fmin, tpe, space_eval
from fastsklearnfeature.candidates.CandidateFeature import CandidateFeature
from fastsklearnfeature.candidates.RawFeature import RawFeature
from fastsklearnfeature.transformations.OneHotTransformation import OneHotTransformation
from typing import List, Dict, Set
from ortools.sat.python import cp_model
from fastsklearnfeature.interactiveAutoML.Runner import Runner
numeric_representations: List[CandidateFeature] = pickle.load(
open("/home/felix/phd/feature_constraints/experiment1/features.p", "rb"))
my_runner = Runner(c=1.0, labels=['bad', 'good'])
filtered = numeric_representations
'''
filtered = []
for f in numeric_representations:
if isinstance(f, RawFeature):
filtered.append(f)
else:
if isinstance(f.transformation, OneHotTransformation):
filtered.append(f)
'''
y_test = pickle.load(
open("/home/felix/phd/feature_constraints/experiment1/y_test.p", "rb"))
open(
"/home/felix/phd/feature_constraints/" + str(which_experiment) +
"/y_train.p", "rb"))
all_features = CandidateFeature(IdentityTransformation(-1),
numeric_representations)
all_standardized = CandidateFeature(MinMaxScalingTransformation(),
[all_features])
foreigner = np.array(X_train[:, 7])
gender = np.array(
['female' in personal_status for personal_status in X_train[:, 15]])
sensitive_feature = gender
my_runner = Runner(c=1.0, sensitive=sensitive_feature, labels=['bad', 'good'])
for c in [0.001, 0.01, 0.1, 1, 10, 100, 1000]:
model = LogisticRegression(penalty='l1',
C=1,
solver='saga',
class_weight='balanced',
max_iter=10000,
multi_class='auto')
my_pipeline = Pipeline([('f', all_standardized.pipeline), ('c', model)])
my_pipeline.fit(X_train, y_train)
sensitive_values = list(np.unique(sensitive_feature))
"/home/felix/phd/feature_constraints/" + str(which_experiment) +
"/X_train.p", "rb"))
#print(X_train[:,7])
#todo: measure TP for each group and add objective
#todo: try misclassification constraint
#todo: restart search
bit2results = {}
foreigner = np.array(X_train[:, 7])
gender = np.array(
['female' in personal_status for personal_status in X_train[:, 15]])
my_runner = Runner(c=1.0, sensitive=gender, labels=['bad', 'good'])
#my_runner = Runner(c=1.0, sensitive=foreigner, labels=['bad', 'good'])
history = []
# define an objective function
def objective(features):
#_, test, pred_test, std_score, proba_pred_test = my_runner.run_pipeline(features, runs=1)
results = my_runner.run_pipeline(features, runs=1)
#print(features)
'''
assert len(y_test) == len(pred_test)
which_observation_should_be_predicted_correctly = 333#131
print(proba_pred_test[which_observation_should_be_predicted_correctly])
filtered.append(f)
numeric_representations = filtered
'''
y_test = pickle.load(
open(
"/home/felix/phd/feature_constraints/" + str(which_experiment) +
"/y_test.p", "rb")).values
#todo: measure TP for each group and add objective
#todo: try misclassification constraint
#todo: restart search
bit2results = {}
my_runner = Runner(c=1.0)
# define an objective function
def objective(features):
score, test, pred_test, std_score, proba_pred_test = my_runner.run_pipeline(
features, runs=1)
#print(features)
assert len(y_test) == len(pred_test)
which_observation_should_be_predicted_correctly = 333 #131
print(proba_pred_test[which_observation_should_be_predicted_correctly])
true_class_index = -1
for c_i in range(len(my_runner.pipeline.classes_)):