def fscore(t,p):
per = precision(t,p)
rec = recall(t,p)
return 2.0 * (per*rec) / (per+rec) if per+rec > 0 else 0
def avgFscore(t,p):
l = len(t)
sum = 0
for i in xrange(l):
sum += fscore(t[i],p[i])
return sum/l
labels = data.getLabelsArray()[:1000]
rawData = data.getDataArray()[:1000]
stPrimerov = len(labels)
bad = data.getBadAttributes(rawData,10)
rawData = data.filterArr(rawData,bad)
k = 10;
print "starting %d fold cross validation" % k
print "number of cases: %d" % len(rawData)
print "number of attributes: %d" % len(rawData[0])
#tolerance = [a/100.0 for a in range(40,50) if a%2 == 0]
#meja = [10,12,14,16,18,20]
#allTests = {}
def fscore(t,p):
per = precision(t,p)
rec = recall(t,p)
return 2.0 * (per*rec) / (per+rec) if per+rec > 0 else 0
def avgFscore(t,p):
l = len(t)
sum = 0
for i in xrange(l):
sum += fscore(t[i],p[i])
return sum/l
labels = data.getLabelsArray()
rawData = data.getDataArray()
stPrimerov = len(labels)
#bad = data.getBadAttributes(rawData,10)
#rawData = data.filterArr(rawData,bad)
k = 10;
print "starting %d fold cross validation" % k
print "number of cases: %d" % len(rawData)
print "number of attributes: %d" % len(rawData[0])
#tolerance = [a/100.0 for a in range(40,50) if a%2 == 0]
#meja = [10,12,14,16,18,20]
#allTests = {}
import matplotlib.pyplot as plot
import data
from sets import Set
from itertools import chain
from collections import Counter
from random import shuffle
rawL = data.getLabelsArray()
rawD = data.getDataArray()
remLD, remLL = data.removeLeastCommonData(rawD, rawL,5)
remMD, remML = data.removeMostCommonData(rawD, rawL,20)
addD , addL = data.addFakeData(rawD, rawL,50)
#expD, expL = data.removeLeastCommonData(rawD, rawL,5)
#expD, expL = data.removeMostCommonData(expD, expL,20)
#expD, expL = data.addFakeData(expD, expL,80)
expD, expL = data.addFakeData(rawD, rawL,80)
expD, expL = data.removeLeastCommonData(expD, expL,5)
expD, expL = data.removeMostCommonData(expD, expL,20)
#plot.hist(list(chain(*addL)),bins=83)
#plot.hist(list(chain(*rawL)),bins=83)
#plot.hist(list(chain(*remLL)),bins=83)
#plot.hist(list(chain(*remML)),bins=83)
#plot.hist(list(chain(*expL)),bins=83)
#plot.xlabel("stevilo oznak")
#plot.ylabel("stevilo primerov")
#plot.show()
#plot.close()
import data
import math
d = data.getDataArray()
t = data.getTestArray()
a10 = data.getBadAttributes(d,10)
d10 = data.filterArr(d,a10)
t10 = data.filterArr(t,a10)
binD = [[int(x>0) for x in i] for i in d10]
binT = [[int(x>0) for x in i] for i in t10]
logD = [[int(math.ceil(math.log(x) if x > 0 else 0)) for x in i] for i in d10]
logT = [[int(math.ceil(math.log(x) if x > 0 else 0)) for x in i] for i in t10]
newD = []
newT = []
for i in range(len(t10)):
newD.append(list(d10[i])+list(binD[i])+list(logD[i]))
newT.append(list(t10[i])+list(binT[i])+list(logT[i]))
f = file("plusBinLogTraingingData.csv","w")
f.write("\n".join(["\t".join([str(x).replace("c","") for x in i]) for i in newD ]))
f.flush()
f.close()
f = file("plusBinLogTestData.csv","w")
f.write("\n".join(["\t".join([str(x).replace("c","") for x in i]) for i in newT ]))
f.flush()
def fscore(t,p): per = precision(t,p) rec = recall(t,p) return 2.0 * (per*rec) / (per+rec) if per+rec > 0 else 0 def avgFscore(t,p): l = len(t) sum = 0 for i in xrange(l): sum += fscore(t[i],p[i]) return sum/l labels = data.getLabelsArray() rawData = data.getDataArray() stPrimerov = len(labels) #bad = data.getBadAttributes(rawData,10) #rawData = data.filterArr(rawData,bad) k = 10; print "starting %d fold cross validation" % k print "number of cases: %d" % len(rawData) print "number of attributes: %d" % len(rawData[0]) aaa = 0 allPred = [] for i in xrange(k):
