In [1]:

    

%load_ext autoreload
%autoreload 2
from app.evaluation import print_boolean_matrix, get_accuracy_and_plot_confusion
import pandas as pd
from app.training import get_undersample_df
from app.classifier import normalize, EnsembleAllNumeric, get_text_pipeline, get_voting_classifier, DescriptionClassifier, ReadmeClassifier, NumericEnsembleClassifier, normalize, EnsembleAllNumeric, keep_useful_features
from app.constants import VALIDATION_DATA_PATH, PROCESSED_DATA_PATH
from sklearn.ensemble import VotingClassifier


    

In [2]:

    

df = get_undersample_df(pd.read_csv(PROCESSED_DATA_PATH))
df.pop("index")
df.pop("Unnamed: 0")
df = normalize(df)
y = df.pop("label")
val_df = normalize(pd.read_csv(VALIDATION_DATA_PATH))
y_val = val_df.pop("label")


    

In [3]:

    

clf = VotingClassifier(estimators=[('description', DescriptionClassifier()),
                                                     ('readme', ReadmeClassifier()),
                                                     ('ensemble', NumericEnsembleClassifier())],
                               voting='soft')
clf.fit(df, y)


    

    
Out[3]:





VotingClassifier(estimators=[('description', <app.classifier.DescriptionClassifier object at 0x7fa34b6191d0>), ('readme', <app.classifier.ReadmeClassifier object at 0x7fa344612f90>), ('ensemble', <app.classifier.NumericEnsembleClassifier object at 0x7fa34b619490>)],
         n_jobs=1, voting='soft', weights=None)

In [4]:

    

predicted = clf.predict(val_df)
print_boolean_matrix(y_val, predicted)


    

    




\begin{table}[h]
\centering
\caption{Boolean Matrix}
\label{boolean_matrix}
\begin{tabular}{|r|r|r|r|r|}
 \hline
Label & Predicted Correctly & Predicted Incorrectly & Precision & Recall \\ \hline
WEB & 2 & 2 & 0.40 & 0.50 \\ \hline
DOCS & 2 & 2 & 0.50 & 0.50 \\ \hline
HW & 5 & 0 & 0.71 & 1.00 \\ \hline
DEV & 8 & 3 & 0.89 & 0.73 \\ \hline
EDU & 2 & 2 & 1.00 & 0.50 \\ \hline
DATA & 2 & 1 & 0.50 & 0.67 \\ \hline
\multicolumn{3}{|l|}{Weighted Average} & 0.72 & 0.68 \\ \hline
\end{tabular}
\end{table}

In [5]:

    

print list(y_val)
print predicted


    

    




['HW', 'HW', 'EDU', 'EDU', 'DEV', 'WEB', 'EDU', 'DOCS', 'DOCS', 'DATA', 'DEV', 'HW', 'DEV', 'DEV', 'HW', 'DEV', 'DOCS', 'DEV', 'EDU', 'DEV', 'DEV', 'WEB', 'HW', 'DATA', 'DATA', 'WEB', 'DEV', 'DEV', 'WEB', 'DEV', 'DOCS']
['HW' 'HW' 'EDU' 'EDU' 'DEV' 'WEB' 'DEV' 'WEB' 'DOCS' 'DOCS' 'DEV' 'HW'
 'DEV' 'DEV' 'HW' 'DEV' 'DOCS' 'DEV' 'WEB' 'DEV' 'DEV' 'WEB' 'HW' 'DATA'
 'DATA' 'DATA' 'DATA' 'WEB' 'HW' 'DOCS' 'HW']

In [6]:

    

get_accuracy_and_plot_confusion(y_val, predicted, clf.classes_)


    

    













    
Out[6]:





0.67741935483870963

In [12]:

    

val_df = pd.read_csv(VALIDATION_DATA_PATH)
val_df["predicted"] = predicted
val_df[["repository", "label", "predicted", "stargazers"]]


    

    
Out[12]:






  

    

      

      
repository
      
label
      
predicted
      
stargazers
    
  
  

    

      
0
      
ga-chicago/wdi5-homework
      
HW
      
HW
      
0.0
    
    

      
1
      
Aggregates/MI_HW2
      
HW
      
HW
      
0.0
    
    

      
2
      
datasciencelabs/2016
      
EDU
      
EDU
      
38.0
    
    

      
3
      
githubteacher/intro-november-2015
      
EDU
      
EDU
      
13.0
    
    

      
4
      
atom/atom
      
DEV
      
DEV
      
31999.0
    
    

      
5
      
jmcglone/jmcglone.github.io
      
WEB
      
WEB
      
69.0
    
    

      
6
      
hpi-swt2-exercise/java-tdd-challenge
      
EDU
      
DEV
      
1.0
    
    

      
7
      
alphagov/performanceplatform-documentation
      
DOCS
      
WEB
      
5.0
    
    

      
8
      
harvesthq/how-to-walkabout
      
DOCS
      
DOCS
      
64.0
    
    

      
9
      
vhf/free-programming-books
      
DATA
      
DOCS
      
69085.0
    
    

      
10
      
d3/d3
      
DEV
      
DEV
      
56332.0
    
    

      
11
      
carlosmn/CoMa-II
      
HW
      
HW
      
0.0
    
    

      
12
      
git/git-scm.com
      
DEV
      
DEV
      
1258.0
    
    

      
13
      
PowerDNS/pdns
      
DEV
      
DEV
      
704.0
    
    

      
14
      
cmrberry/cs6300-git-practice
      
HW
      
HW
      
0.0
    
    

      
15
      
Sefaria/Sefaria-Project
      
DEV
      
DEV
      
134.0
    
    

      
16
      
mongodb/docs
      
DOCS
      
DOCS
      
445.0
    
    

      
17
      
sindresorhus/eslint-config-xo
      
DEV
      
DEV
      
67.0
    
    

      
18
      
e-books/backbone.en.douceur
      
EDU
      
WEB
      
101.0
    
    

      
19
      
erikflowers/weather-icons
      
DEV
      
DEV
      
3737.0
    
    

      
20
      
tensorflow/tensorflow
      
DEV
      
DEV
      
35812.0
    
    

      
21
      
cs231n/cs231n.github.io
      
WEB
      
WEB
      
1632.0
    
    

      
22
      
m2mtech/smashtag-2015
      
HW
      
HW
      
19.0
    
    

      
23
      
openaddresses/openaddresses
      
DATA
      
DATA
      
754.0
    
    

      
24
      
benbalter/congressional-districts
      
DATA
      
DATA
      
94.0
    
    

      
25
      
Chicago/food-inspections-evaluation
      
WEB
      
DATA
      
131.0
    
    

      
26
      
OpenInstitute/OpenDuka
      
DEV
      
DATA
      
40.0
    
    

      
27
      
torvalds/linux
      
DEV
      
WEB
      
38443.0
    
    

      
28
      
bhuga/bhuga.net
      
WEB
      
HW
      
1.0
    
    

      
29
      
macloo/just_enough_code
      
DEV
      
DOCS
      
1.0
    
    

      
30
      
hughperkins/howto-jenkins-ssl
      
DOCS
      
HW
      
25.0
    
  

In [ ]: