def learningAddIndicateFunctionlfm(n, alpha, lamd):
global fileNameAllPuiArray
p, q = initTrainMatrix()
puiArray = paper.load(fileNameAllPuiArray)
#0-1
puiArray = (puiArray - 44.034)/ 107.439
import random
for step in range(0, n):
number = 0
mae = 0
userList = range(userNum)
while len(userList) > 0:
u = random.choice(userList)
del userList[userList.index(u)]
itemList= range(itemNum)
while len(itemList) > 0:
i = random.choice(itemList)
del itemList[itemList.index(i)]
if puiArray[u, i] != 111:
number += 1 #number 记录训练集中的评分总数
rui = puiArray[u, i]
pui = predict(u, i, p, q)
eui = rui - pui
mae += abs(eui)
#修正 p\q
temp = p[u,] + alpha * (q[i, ] * eui - lamd * p[u,])
q[i,] += alpha * (p[u,] * eui - lamd * q[i,])
p[u,] = temp
else:
p[u,] -= alpha * lamd * p[u,]
q[i,] -= alpha * lamd * q[i,]
nowMae = mae / number
print 'step: %d mae: %f' % ((step + 1), nowMae)
alpha *= 0.9
return p, q
def loadPQ(*fileName):
import paper
p = paper.load(fileName[0]).A
q = paper.load(fileName[1]).A
return p, q
import sys
import copy
if __name__ == '__main__':
simCalMethod = paper.simMinkowskiDist
sparseness = 'sparseness5'
fileNumbers = 1
K = 2
MAE = 0.0
RMSE = 0.0
#参数调整 0.519026063656 1.64540380436
#文件对象
for i in range(1, fileNumbers+1):
#文件对象
trainFileName = 'dataSet/%s/training%d.txt' % (sparseness,i)
testFileName = 'dataSet/%s/test%d.txt' % (sparseness,i)
euiFileName = 'result/eui/euiNR-%s-test%d.txt' % (sparseness,i)
pf = open(euiFileName, 'w')
#load data
testDict = paper.loadTestDataForDict(testFileName)
trainArrayObj = paper.createArrayObj(trainFileName)
mean = paper.calMean(trainArrayObj)
wsAverageVector = paper.columnAverageArray(trainArrayObj, mean)
#相似度矩阵数据
userSimFileName = 'result/simArraySlopeoneUser-%s-test%d.txt' % (sparseness,i)
# userSimArrayObj = paper.createSimArray(trainArrayNorm, simCalMethod, 2.0)
# paper.save(userSimArrayObj, userSimFileName)
userSimArrayObj = paper.load(userSimFileName)
#计算预测准确
print calMaeAndRmse()
pf.close()
print 'ok'
def loadPQTp(sparseness=5, number=1):
pFile = r'throught/p-%d-%d.txt' % (sparseness, number)
qFile = r'throught/q-%d-%d.txt' % (sparseness, number)
p = paper.load(pFile)
q = paper.load(qFile)
return p, q