def svm_process(self,theList) :
# SVM process
np.set_printoptions(threshold=np.NaN)
datas, labels = self.loadData(theList) #dalam datas yang berbentuk list ada dict
print("sebelum normalisasi")
#win.textbox.insertPlainText("sebelum normalisasi")
#win.add_table_data(datas)
#desktop2.add_text(desktop.window,"jancokqq")
datas = svm.normalize(datas,win.normType.currentText())
print("setelah normalisasi")
#win.add_table_data2(datas)
list_temp=list()
for i in datas:
sublist=list()
for key in i :
sublist.append(i[key])
list_temp.append(sublist)
datas=mat(list_temp)
m, n = shape(datas)
kernel= mat(zeros((m,m)))
print("size=",m)
for i in range(m) :
kernel[:,i] = svm.kernelTrans(datas, datas[i, :],kTup={'type': win.kernelType.currentText(), 'gamma': win.paramGamma.value(),'degree': win.paramDegree.value()\
,'coef' :win.paramCoef.value()})
#print("kernel=",kernel)
b, alphas = svm.smoSimple(kernel, labels, win.paramC.value(), win.paramTolerance.value(), win.paramIterate.value())
# print('alphas')
# print(alphas)
# print('b')
# print(b)
return b, alphas
def eval(self,alphas,b,datas,training) :
dataTrain,labelTrain = self.loadData(training)
dataTrain = svm.normalize(dataTrain,win.normType.currentText())
dataEval, labelEval = self.loadData(datas)
dataEval = svm.normalize(dataEval,win.normType.currentText())
dataTrain=mat(self._dictToList(dataTrain))
dataEval=mat(self._dictToList(dataEval))
#dataEval = mat(dataEval)
m, n = shape(dataEval)
labelTrain = mat(labelTrain).transpose()
panjang = list() #simpan data buat f-score
errorCount = 0
for i in range (m) :
kernelEval = svm.kernelTrans(dataTrain, dataEval[i, :], kTup={'type': win.kernelType.currentText(), 'gamma': win.paramGamma.value(),'degree': win.paramDegree.value()\
,'coef' :win.paramCoef.value()})
predict = kernelEval.T * multiply(labelTrain, alphas) + b
panjang.append(float(sign(predict)))
#print("panjang=",predict)
if sign(predict) != sign(labelEval[i]): errorCount += 1
print("the error rate is: %f" % (float(+errorCount) / m))
print("labelArr", labelEval)
print("panjang", panjang)
f1_positive = f1_score(labelEval, panjang,pos_label=1)
f1_negative = f1_score(labelEval, panjang, pos_label=-1)
print("the f1_score is :", (f1_positive + f1_negative) / 2)
win.label_fscore.setText("f-score :"+str((f1_positive + f1_negative) / 2))
def testPrc(kTup=('rbf', 10)):
dataArr, labelArr = loadDataSet('201404')
print 'load data done'
b, alphas = smoP(dataArr, labelArr, 200, 0.0001, 10000, kTup)
datMat = mat(dataArr);
labelMat = mat(labelArr).transpose()
svInd = nonzero(alphas.A > 0)[0]
sVs = datMat[svInd]
labelSV = labelMat[svInd];
print "there are %d Support Vectors" % shape(sVs)[0]
saveSVM(svInd, sVs, labelSV, alphas[svInd]) #保存支持向量
m, n = shape(datMat)
errorCount = 0
for i in range(m):
kernelEval = kernelTrans(sVs, datMat[i, :], kTup)
predict = kernelEval.T * multiply(labelSV, alphas[svInd]) + b
if sign(predict) != sign(labelArr[i]): errorCount += 1
print "the training error rate is: %f" % (float(errorCount) / m)
def handwritingIndentify(kTup):
featureList,labelList = loadImage('digits/trainingDigits')
print featureList
print labelList
b, alphas = svm.smoP(array(featureList),labelList,200,0.0001,10000,kTup)
print "b:",b, "alpha:",alphas
svInd = nonzero(alphas.A > 0)[0]
sVs = mat(array(featureList))[svInd]
labelSV= mat(labelList).T[svInd]
print "there a %s Support Vectors."%shape(sVs)[0]
featureMat = mat(array(featureList));labelMat = mat(labelList).T
m,n = shape(featureMat)
errorCount = 0
for i in range(m):
kernelMat = svm.kernelTrans(sVs, featureMat[i,:], kTup)
predict = kernelMat.T * multiply(labelSV,alphas[svInd]) + b
if sign(predict) != sign(labelMat[i]):
errorCount += 1
print "the training data error rate is %s"%(float(errorCount)/m)
def handwritingIndentify(kTup):
featureList, labelList = loadImage('digits/trainingDigits')
print featureList
print labelList
b, alphas = svm.smoP(array(featureList), labelList, 200, 0.0001, 10000,
kTup)
print "b:", b, "alpha:", alphas
svInd = nonzero(alphas.A > 0)[0]
sVs = mat(array(featureList))[svInd]
labelSV = mat(labelList).T[svInd]
print "there a %s Support Vectors." % shape(sVs)[0]
featureMat = mat(array(featureList))
labelMat = mat(labelList).T
m, n = shape(featureMat)
errorCount = 0
for i in range(m):
kernelMat = svm.kernelTrans(sVs, featureMat[i, :], kTup)
predict = kernelMat.T * multiply(labelSV, alphas[svInd]) + b
if sign(predict) != sign(labelMat[i]):
errorCount += 1
print "the training data error rate is %s" % (float(errorCount) / m)
file = "C:\\Users\\Daniel\\PycharmProjects\\Spotiwew\\Sandro Yusuf Sihaloho\\main.csv"
csvReader = csv.DictReader(open(file))
thelist = list()
for i in csvReader:
thelist.append(i)
datas, labels = loadData(
thelist
) #dalam datas yang berbentuk list ada dict. disini memisahkan data dan label
#print(thelist)
list_temp = list()
for i in datas:
sublist = list()
for key in i:
sublist.append(i[key]) # get value of the key
list_temp.append(sublist) # add to list
datas = mat(list_temp)
m, n = shape(datas)
kernel = mat(zeros((m, m)))
print("size=", m)
for i in range(m):
kernel[:, i] = svm.kernelTrans(datas,
datas[i, :],
kTup={
'type': "laplacian",
'gamma': 0.5
})
# print("kernel=",kernel)
b, alphas = svm.smoSimple(kernel, labels, C=10, toler=0.1, maxIter=2)
print(b)
print(alphas)
def recommend(self) :
with open("resultSVM.txt", "rb") as fp:
alphas, b, training = pickle.load(fp)
connect.set_headers(win.labelbox.text())
song=list()
with open("listPositive.txt", "rb") as fp:
listPositive= pickle.load(fp)
print("listPositive=",listPositive)
listPositiveNew = list()
if len(listPositive)>200 : #biar tidak tlalu banyak request]
listPositive=self.random_del(listPositive,200)
print("listnew=", len(listPositive))
if connect.get_display_name() != None:
mypath = str(connect.get_display_name())
else:
mypath = str(connect.get_current_user())
for i in listPositive :
tempJson=connect.recommendation_based_seeds(i)
for j in tempJson['tracks'] :
song.append(j['id'])
if not os.path.exists(mypath+"/recommendations.csv"):
print("cari rekomendasi")
headers = ['id', 'danceability', 'energy', 'loudness', 'speechiness', 'acousticness', 'instrumentalness', 'valence',
'tempo']
self.create_csv("recommendations.csv",mypath,headers)
mypath = mypath+"/recommendations.csv"
print(mypath)
self.get_audio_features(song,mypath) #skaligus write to csv
else :
mypath = mypath + "/recommendations.csv"
recomTrain, labelTrain = self.loadData(training)
# win.add_table_data(dataTrain)
recomTrain = svm.normalize(recomTrain, win.normType.currentText())
templist=list()
for i in recomTrain :
sublist=list()
for key in i :
sublist.append(i[key])
templist.append(sublist)
recomTrain=templist
print("dataTrain=", shape(mat(recomTrain)))
print(recomTrain)
# win.add_table_data2(dataTrain)
# win.addAlphasItem(alphas)
# win.label_alphas.setText("alphas size :" + str(len(alphas)))
# win.label_bias.setText("Bias :"+str(b))
csvReader = csv.DictReader(open(mypath))
RecoCSV=list()
print(csvReader)
for i in csvReader :
RecoCSV.append(i)
dataReco, id_lagu = self.loadData(RecoCSV) # simpan id lagu bukan label
print("datareco=",dataReco)
dataReco = svm.normalize(dataReco, win.normType.currentText())
print("datareconorm=", dataReco)
templist2=list()
for j in dataReco :
sublist2=list()
for key2 in j :
sublist2.append(j[key2])
#print("sublist2=",sublist2)
templist2.append(sublist2)
dataReco=templist2
print(dataReco)
recomTrain = mat(recomTrain)
print(recomTrain)
#win.add_table_dataRecom(dataReco)
dataReco = mat(dataReco)
m, n = shape(dataReco)
labelTrain = mat(labelTrain).transpose()
print('udah mo abis bgst')
print(id_lagu)
listPredict=list()
print("komall", shape(recomTrain))
print(shape(dataReco))
print(shape(labelTrain))
print(shape(alphas))
print("datareco=",dataReco)
print("dataTrain=", recomTrain)
for i in range(m):
if (id_lagu[i] in listPredict)==False:
kernelEval = svm.kernelTrans(recomTrain, dataReco[i, :],
kTup={'type': win.kernelType.currentText(), 'gamma': win.paramGamma.value(),
'degree': win.paramDegree.value(), 'coef': win.paramCoef.value()})
predict = kernelEval.T * multiply(labelTrain, alphas) + b
temp = {'id' : id_lagu[i],"value":predict }
if temp not in listPredict :
listPredict.append(temp)
print("list predict=",listPredict)
print("list predict=", type(listPredict))
seen=set()
newListPredict=list()
# for d in list(listPredict) : #in here we delete the dict with same
# t=tuple(d.items())
# if t not in seen :
# seen.add(t)
# newListPredict.append(t)
# listPredict=newListPredict
listPredict.sort(key=itemgetter('value'),reverse=True)
print("listPredict=",listPredict)
# listRecomm = list()
# count=0
# while len(listRecomm) < 10:
# temp_l.append(listPredict[count])
# count+=1
listTemp=listPredict[:10]
listRecomm=list()
for i in listTemp :
listRecomm.append(i['id'])
recom_song = self.__list_to_str(listRecomm)
print(recom_song)
result = connect.get_several_tracks(recom_song)
self.get_song_name(result)