dclambert / pyensemble Goto Github PK

Hey! Great work on simplifying the process of building ensembles. I was trying to use this, but I found that the GBM implementation of Scikit learn is far too slow for my needs. We could possibly include a faster implementation, aka XGBoost.

I would love to help out with this, will make a PR soon after adding XGBoost. Also keen on helping you add more scoring functions (precision, recall, matthew's coefficient, and support for custom scoring functions similar to the scikit learn api)

Cheers!

Not sure if you were still maintaining this code base base but now fails with sklearn 0.16.0 due to:

"The min_density parameter is deprecated as of version 0.14 and will be removed in 0.16.
"and will be removed in 0.16.", DeprecationWarning)"

Hi:

I tested the simplest call of ensemble_train and got a ValueError for the parameter min_samples_split:

Traceback (most recent call last):
File "pyensemble/ensemble_train.py", line 202, in
ens.fit(X_train, y_train)
File "/home/mourao/income_prediction/pyensemble/ensemble.py", line 290, in fit
self.fit_models(X, y)
File "/home/mourao/income_prediction/pyensemble/ensemble.py", line 325, in fit_models
model.fit(X[train_inds], y[train_inds])
File "/usr/local/lib/python2.7/dist-packages/sklearn/tree/tree.py", line 790, in fit
X_idx_sorted=X_idx_sorted)
File "/usr/local/lib/python2.7/dist-packages/sklearn/tree/tree.py", line 194, in fit
% self.min_samples_split)
ValueError: min_samples_split must be an integer greater than 1 or a float in (0.0, 1.0]; got the integer 1

I solved the problem removing 1 from the list in the file model_library.py:

def build_decisionTreeClassifiers(random_state=None):
    rs = check_random_state(random_state)

    param_grid = {
        'criterion': ['gini', 'entropy'],
        'max_features': [None, 'auto', 'sqrt', 'log2'],
        'max_depth': [None, 1, 2, 5, 10],
        'min_samples_split': [2, 5, 10],
        'random_state': [rs.random_integers(100000) for i in xrange(3)],
    }

    return build_models(DecisionTreeClassifier, param_grid)