Gizmomens

Hi, thanks for the response! I've attached the model and the data set, the original scikit model on python was giving me an accuracy of ~70%. [model_data.zip](https://github.com/nok/sklearn-porter/files/1670225/model_data.zip)

So the low accuracy in the exported Java model is because of these bugs then?

But the accuracy is bad even if I use the same test data in Java. The scikit model gives 70% on that test set while the Java code gives me...

This can't be right. The Java code also returns the classes "2" and "1" most of the time just as shown above but the data setis almost equally distributed in...

I see, but what I don't understand is, if the estimator is that weak, how was scikit giving me such a good result? ``` RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini', max_depth=None, max_features='auto', max_leaf_nodes=None,...

``` C:\Python\Scripts>pip list Message: 'beautifulsoup4 (4.6.0)' Arguments: () beautifulsoup4 (4.6.0)bleach (1.5.0) enum34 (1.1.6) html5lib (0.9999999) Markdown (2.6.11) numpy (1.14.0) pip (8.1.1) protobuf (3.5.1) scikit-learn (0.19.1) scipy (1.0.0) setuptools (20.10.1) six...