def maeAndRmseTp(p, q, testArrayObj, pfEui):
euiList = []
for index in range(len(testArrayObj)):
u = int(testArrayObj[index, 0])
i = int(testArrayObj[index, 1])
tui = testArrayObj[index, 2]
pui = (p[u] * q[i].T)[0, 0] * 107.439 + 44.034
euiTemp = tui - pui
euiList.append(euiTemp)
pfEui.write(str(euiTemp) + '\n')
paper.maeAndRmse(euiList)
def predWithPQTp(testDataArray, pf, p, q):
euiList = []
for index in range(len(testDataArray)):
u = int(testDataArray[index, 0])
i = int(testDataArray[index, 1])
tui = testDataArray[index, 2]
pui = (p[u] * q[i].T)[0, 0] * 107.439 + 44.034
euiTemp = tui - pui
pf.write(str(euiTemp) + '\n')
euiList.append(euiTemp)
paper.maeAndRmse(euiList)
def test1(self):
import numpy as np
fileIndex = 1
for i in xrange(1):
# userEui = np.loadtxt(ROOT_DIRECTORY + "/userEui" + str(fileIndex) + ".txt")
userEui = np.loadtxt(ROOT_DIRECTORY + "/userEui.txt")
# wsEui = np.loadtxt(ROOT_DIRECTORY + "/wsEui" + str(fileIndex) + ".txt")
wsEui = np.loadtxt(ROOT_DIRECTORY + "/wsEui.txt")
fileIndex = fileIndex + 1
alpha = 0.4
count = 0
totalQos = 0
testArr = paper.loadTest(ROOT_DIRECTORY + '/test1.txt')
minMae = 1000
minAlpha = 10
for i in xrange(11):
eui = []
alpha = i / 10.0
totalQos = 0
for index, value in enumerate(userEui):
totalQos = totalQos + testArr[index][2]
eui.append(testArr[index][2] - alpha * value -
(1 - alpha) * wsEui[index])
eui = np.array(eui)
mae, rmse = paper.maeAndRmse(eui)
if mae < minMae:
minMae = mae
minAlpha = alpha
print "alpha:" + str(alpha) + "\t" + "MAE:" + str(
mae) + "\tRMSE:" + str(rmse) + "\tNMAE:" + str(
totalQos / len(testArr))
print "最小mae:" + str(minMae) + "\t alpha=" + str(minAlpha)
def calAverageMaeAndRmse(start=1, end=10, sparseness=5):
maeSums = rmseSums = 0.0
for i in range(start, end + 1):
# userEuiFile = r'rt/eui/slopeone-%d-%d.txt' % (sparseness,i)
userEuiFile = r'throught/eui/slopeone-%d-%d.txt' % (sparseness, i)
result = np.loadtxt(userEuiFile)
mae, rmse = paper.maeAndRmse(result)
maeSums += mae
rmseSums += rmse
print maeSums / (end - start + 1), rmseSums / (end - start + 1)
def findBestMatch(a, b):
lamda = 0.0
minMae = sys.maxint
minRmse = sys.maxint
minLamda = -1
while lamda <= 1.0:
result = a * lamda + b * (1 - lamda)
print lamda,
tempMae, tempRmse = paper.maeAndRmse(result)
if tempMae < minMae:
minMae = tempMae
minRmse = tempRmse
minLamda = lamda
lamda += 0.1
print minLamda, minMae, minRmse
def test1(self):
import numpy as np
trainingArr = createArrayObj(ROOT_DIRECTORY + '/training1.txt', 339,
5825)
simUserArr = np.loadtxt(ROOT_DIRECTORY + "/simUserArr.txt")
simWsArr = np.loadtxt(ROOT_DIRECTORY + "/simWsArr.txt")
webServerMapping = np.loadtxt(ROOT_DIRECTORY + "/webServerMapping.txt",
dtype=str,
delimiter="\t")
maxSimUser = np.argsort(simUserArr, axis=1)[:, 0 - 10:]
testArr = paper.loadTest(ROOT_DIRECTORY + '/test1.txt')
#标签
pui = np.loadtxt(ROOT_DIRECTORY + "/label.txt")
#标签－概率
probability = np.loadtxt(ROOT_DIRECTORY + "/probability.txt")
# 取出相似度最高的k个用户
k_user = 10
k_ws = 10
result = []
fileIndex = 1
for i in xrange(1):
k_user = 10 + i * 10
for j in xrange(1):
k_ws = 50 + j * 10
simUser = np.argsort(simUserArr, axis=1)[:, 0 - k_user:]
simWs = np.argsort(simWsArr, axis=1)[:, 0 - k_ws:]
eui = []
# 用户－服务
for index, value in enumerate(testArr):
uId = value[0]
wsId = value[1]
label = pui[index]
fenzi = 0
fenmu = 0
# 相似用户－相似度
for i, otherUId in enumerate(simUser[uId]):
#pow(simUserArr[uId][otherUId], 3)
#simUserArr[uId][otherUId]
if trainingArr[otherUId][wsId] != NoneValue and int(
trainingArr[otherUId][wsId]) in label:
fenzi += simUserArr[uId][otherUId] * trainingArr[
otherUId][wsId] * probability[otherUId][
np.argwhere(label == int(
trainingArr[otherUId][wsId]))[0, 0]]
fenmu += simUserArr[uId][otherUId] * probability[
otherUId][np.argwhere(label == int(
trainingArr[otherUId][wsId]))[0, 0]]
if fenmu != 0:
predictValue = fenzi / fenmu
r = predictValue - value[2]
if r < 0:
r = 0 - r
eui.append(r)
else:
total = 0
count = 0
for i, otherUId in enumerate(maxSimUser[uId]):
for m, n in enumerate(simWs[wsId]):
# if webServerMapping[wsId][2] != webServerMapping[n][2] and webServerMapping[wsId][4] != webServerMapping[n][4]:
# continue;
#maxSimUser[uId][0]
if trainingArr[otherUId][n] != NoneValue:
count = count + 1
total += trainingArr[otherUId][n]
if count != 0:
avg = total / count
if avg < 0:
avg = 0 - avg
eui.append(avg)
print len(eui)
eui = np.array(eui)
mae, rmse = paper.maeAndRmse(eui)
result.append("相似用户数:" + str(k_user) + "\t相似服务数:" + str(k_ws) +
"\t" + "MAE:" + str(mae) + "\tRMSE:" + str(rmse))
print "相似用户数:" + str(k_user) + "\t相似服务数:" + str(
k_ws) + "\t" + "MAE:" + str(mae) + "\tRMSE:" + str(rmse)
# np.savetxt(ROOT_DIRECTORY + "/result.txt", result, fmt='%s', delimiter="\t")
np.savetxt(ROOT_DIRECTORY + "/userEui.txt",
eui,
fmt='%s',
delimiter="\t")
# np.savetxt(ROOT_DIRECTORY + "/userEui" + str(fileIndex) + ".txt", eui, fmt='%s', delimiter="\t")
fileIndex = fileIndex + 1
def getMaeAndRmse(a, b, lamda):
result = a * lamda + b * (1 - lamda)
mae, rmse = paper.maeAndRmse(result)
return mae, rmse
def test1(self):
import numpy as np
trainingArr = createArrayObj(ROOT_DIRECTORY + '/training1.txt', 339,
5825)
simArr = createSimArray(trainingArr, paper.simMinkowskiDist, 2)
webServerMapping = np.loadtxt(ROOT_DIRECTORY + "/webServerMapping.txt",
dtype=str,
delimiter="\t")
# 取出相似度最高的k个用户
k = 10
simUser = np.argsort(simArr, axis=1)[:, 0 - k:]
testArr = paper.loadTest(ROOT_DIRECTORY + '/test1.txt')
pui = np.loadtxt(ROOT_DIRECTORY + "/label.txt")
probability = np.loadtxt(ROOT_DIRECTORY + "/probability.txt")
eui = []
# 用户－服务
for index, value in enumerate(testArr):
uId = value[0]
wsId = value[1]
label = pui[index]
fenzi = 0
fenmu = 0
# 相似用户－相似度
for i, otherUId in enumerate(simUser[uId]):
if trainingArr[otherUId][wsId] != NoneValue and int(
trainingArr[otherUId][wsId]) in label:
fenzi += simArr[uId][otherUId] * trainingArr[otherUId][
wsId] * probability[otherUId][np.argwhere(
label == int(trainingArr[otherUId][wsId]))[0, 0]]
fenmu += simArr[uId][otherUId] * probability[otherUId][
np.argwhere(
label == int(trainingArr[otherUId][wsId]))[0, 0]]
if fenmu != 0:
predictValue = fenzi / fenmu
eui.append(predictValue - value[2])
# else:
# if webServerMapping[wsId - 1][2] != 'not found':
# ii = 2
# else:
# ii = 4
# score = [ ]
# for g, h in enumerate(webServerMapping):
# if h[ii] == webServerMapping[wsId - 1][ii]:
# # 取出和i同个as的其他服务
# score.append(int(h[0]))
# total = 0
# count = 0
# for i, otherUId in enumerate(simUser[uId]):
# for m, n in enumerate(score):
# if trainingArr[otherUId][n] != NoneValue:
# count = count + 1
# total += trainingArr[otherUId][n]
# if count != 0:
# avg = total / count;
# eui.append(avg)
print len(eui)
eui = np.array(eui)
mae, rmse = paper.maeAndRmse(eui)
print "MAE:" + str(mae) + "\tRMSE:" + str(rmse)
if not (int(info[-5]), int(info[-4])) in temp:
result.append(info)
result = sorted(result, key=lambda x: float(x[-1]))
with open('euiAnalysis/euiFilter1.txt','w') as pf:
for line in result:
for elem in line:
pf.write(str(elem) + '\t')
pf.write('\n')
def filters():
eui_c1 = r'euiAnalysis/euiWith-1-c1.txt'
eui_NotGet = r'euiAnalysis/euiNotGet.txt'
result = []
with open(eui_c1) as pf:
for line in pf:
info = line.strip().split()
u = int(info[-5])
i = int(info[-4])
result.append((u, i))
with open(eui_NotGet) as pf:
for line in pf:
info = line.strip().split()
u = int(info[-5])
i = int(info[-4])
result.append((u, i))
euiFilter(result)
if __name__ == '__main__':
paper.maeAndRmse([-1, 1, 2])