#from support import load_bulldozer
import support
from stratx.partdep import plot_stratpd, plot_catstratpd, plot_catstratpd_gridsearch
from stratx.featimp import importances
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
np.set_printoptions(precision=2, suppress=True, linewidth=300, threshold=2000)
#np.random.seed(1)
# n = 25_000
# X, y = support.load_bulldozer(n)
X, y, X_train, X_test, y_train, y_test = support.load_dataset("bulldozer", "SalePrice")
X['auctioneerID'] = X['auctioneerID'].astype(np.int64)
# df = pd.read_csv("../../pd/bulldozer20k.csv")
# X = df.drop('SalePrice', axis=1)
# y = df['SalePrice']
I = importances(#X, y,
X_train, y_train,
n_trials=1,
normalize=False,
drop_high_stddev=2.0,
min_samples_leaf=20,
cat_min_samples_leaf=20,
from sklearn.utils import resample
from sklearn.model_selection import train_test_split
from timeit import default_timer as timer
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import GridSearchCV
import xgboost as xgb
from sklearn import svm
np.random.seed(1) # choose a seed that demonstrates diff RF/GBM importances
# boston = load_boston()
# X = pd.DataFrame(boston.data, columns=boston.feature_names)
# y = pd.Series(boston.target)
X, y, X_train, X_test, y_train, y_test = load_dataset("boston", "MEDV")
n = X.shape[0]
n_shap = len(X_test) # test all
fig, axes = plt.subplots(1, 4, figsize=(10, 2.5))
lm = LinearRegression()
X_train_ = StandardScaler().fit_transform(X_train)
X_train_ = pd.DataFrame(X_train_, columns=X.columns)
X_test_ = StandardScaler().fit_transform(X_test)
X_test_ = pd.DataFrame(X_test_, columns=X.columns)
lm.fit(X_train_, y_train)
lm_score = lm.score(X_test_, y_test)
print("OLS validation R^2", lm_score)
ols_shap_I = shap_importances(lm, X_train_, X_test_,
import support
from stratx.partdep import plot_stratpd, plot_catstratpd,\
plot_catstratpd_gridsearch, plot_stratpd_gridsearch
import matplotlib.pyplot as plt
from stratx.featimp import importances
import numpy as np
from sklearn.model_selection import train_test_split
from sklearn.ensemble import RandomForestRegressor
#np.random.seed(5)
X, y, X_train, X_test, y_train, y_test = support.load_dataset("rent", 'price')
# X, y = X[:100], y[:100]
# X, y = support.load_rent(n=15_000)
# print(X.shape)
# X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2)
# tuned_params = support.models[("rent", "RF")]
# rf = RandomForestRegressor(**tuned_params, n_jobs=-1)
# rf.fit(X_train, y_train)
# print("R^2 test",rf.score(X_test,y_test))
I = importances(
X,
y,
n_trials=1,
normalize=False,
bootstrap=True,
# bootstrap=False,
# subsample_size=.7,
import support
from stratx import importances, plot_importances
import matplotlib.pyplot as plt
import numpy as np
from support import synthetic_files
for dataset in synthetic_files:
print(f"Plot stability for {dataset}")
np.random.seed(1)
X, y, X_train, X_test, y_train, y_test = support.load_dataset(
dataset, targetname='response')
print(X.shape)
I = importances(X,
y,
bootstrap=False,
n_trials=30,
subsample_size=.75,
n_jobs=4)
print(I)
# Don't need for continuous x_i since importance == impact when
# there are n unique values for n values
# plot_importances(I[0:10], imp_range=(0, 0.4), sortby='Importance')
#plt.savefig(f"../images/{dataset}-stability-importance.pdf", bbox_inches="tight", pad_inches=0)
# plt.show()
# plt.close()
plot_importances(I[0:20], imp_range=(0, 0.4), sortby='Impact')