#!/usr/bin/env python
import utils
if __name__ == '__main__':
import classifiers
from datamodel import *
import database as db
s=db.Session()
evaluated=s.query(Link).filter(Link.evaluation != None).count()
links=s.query(Link).filter(Link.evaluation != None).order_by(Link.date).all()[::2]
print "Training on %d of %d evaluated links (every second element)" % (len(links),evaluated)
for cl_name in utils.get_classifiers():
current=__import__('classifiers.'+cl_name,fromlist=[classifiers])
print "Training",cl_name
current.train(links)
def show_classifiers():
import classifiers
for cl_name in utils.get_classifiers():
current=__import__('classifiers.'+cl_name,fromlist=[classifiers])
print "- "+cl_name+":";current.print_self()
def test_classifiers():
from math import fabs
cl_evaluations={}
for cl_name in utils.get_classifiers():
cl_evaluations[cl_name]=ClassifierEvaluation(cl_name)
return cl_evaluations
def compute(shift):
before_after = pandas.DataFrame()
for key in list(itertools.product(utils.get_projects(), utils.get_class_models(), utils.get_experiments(), ['ideal'], utils.get_smote(), utils.get_approaches(), utils.get_classifiers())):
data = utils.get_item(key, 'data')
before_after_local = compute_before_after_series(data, shift)
before_after_local['project'] = key[0]
before_after_local['approach'] = key[5]
before_after = before_after.append(before_after_local, sort=False)
utils.boxplot_approaches(before_after, 'before_after_fix_' + str(shift), 'difference', '', True, [-1, 1])
