wehub-resource-sync
27 lines
743 B
Python
27 lines
743 B
Python
# Create 'crime.pickle'
|
|
|
|
import pickle
|
|
|
|
import lightgbm as lgb
|
|
from sklearn.model_selection import train_test_split
|
|
|
|
import shap
|
|
|
|
random_state = 1203344
|
|
|
|
# Load data and train model
|
|
X, y = shap.datasets.communitiesandcrime()
|
|
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=random_state)
|
|
model = lgb.LGBMRegressor(random_state=random_state)
|
|
model.fit(X_train, y_train, eval_set=[(X_test, y_test)], verbose=False)
|
|
|
|
# Calculate and plot SHAP values
|
|
explainer = shap.TreeExplainer(model)
|
|
idx = 13
|
|
shap_values = explainer.shap_values(X_test.iloc[[idx]], y_test[idx])
|
|
|
|
# Dump to pickle
|
|
o = (explainer.expected_value, shap_values, X_test.iloc[0])
|
|
with open("./crime.pickle", "wb") as fl:
|
|
pickle.dump(o, fl)
|