def feature_importance(boost_obj, args):
plt.figure(figsize=[10, 10])
model = args.model
path = args.out
if model.lower() == 'gboost':
import xgboost as xgb
xgb.plot_importance(boost_obj.model, color='green')
plt.rcParams['figure.figsize'] = [10, 10]
'''
xgb.plot_tree(boost_obj.model, num_trees=0, rankdir='LR')
fig = matplotlib.pyplot.gcf()
fig.set_size_inches(150, 100)
fig.savefig('tree.png')
'''
elif model.lower() in ['lr', 'svml']:
pd.Series(abs(boost_obj.model.coef_[0]),
index=boost_obj.X_data.columns).plot(kind='barh')
elif model.lower() in ['ngboost', 'svm_rbf']:
return 0
else:
importances_rf = pd.Series(boost_obj.model.feature_importances_,
index=boost_obj.X_data.columns)
sorted_importance_rf = importances_rf.sort_values()
sorted_importance_rf.plot(kind='barh', color='lightgreen')
plt.title('Feature Importance')
if args.snps:
name = 'feature_importance_' + args.predictor + '_snps.pdf'
elif args.indels:
name = 'feature_importance_' + args.predictor + '_indels.pdf'
elif args.all:
name = 'feature_importance_' + args.predictor + '.pdf'
plt.savefig(os.path.join(path, name), bbox_inches='tight')
def feature_importance(boost_obj, model, path):
plt.figure(figsize=[10, 10])
if model.lower() == 'gboost':
import xgboost as xgb
xgb.plot_importance(boost_obj.model)
plt.rcParams['figure.figsize'] = [10, 10]
else:
importances_rf = pd.Series(boost_obj.model.feature_importances_,
index=boost_obj.X_data.columns)
sorted_importance_rf = importances_rf.sort_values()
sorted_importance_rf.plot(kind='barh', color='lightgreen')
plt.title('Feature Importance')
plt.savefig(os.path.join(path, 'feature_importance.png'))
