def run_para(oriTrain,oriTest,field,numField,postfix,isToCSV=True):
if isToCSV: csv_2_ffm.to_ffm(oriTrain,oriTest,field,numField=numField,postfix= postfix)
trainFile = oriTrain[0:-4]+ '_'+ postfix +'.ffm'
testFile = oriTest[0:-4]+ '_'+ postfix +'.ffm'
t=str(150) ; k=str(8) ; l=str(0.00002)
resultName = '../submit/ffm_'+ postfix + '_t{t}_k{k}_l{l}'.format(t=t,k=k,l=l,feat=str(len(field)))
assignCMDPara=' -t {t} -k {k} -l {l} -s 4 -r 0.02 '.format(t=t,k=k,l=l)
ffm.runffm(trainFile,testFile,resultName,assignCMDPara)
def onceRun(day):
oriTrain = '../data/temp/train_preAction_{day}.csv'.format(day=day)
oriTest = '../data/temp/test_preAction_19.csv'
postfix='m_p_u_amc_s{feat}_preAction_{day}'.format(feat=str(len(field)),day=day)
csv_2_ffm.to_ffm(oriTrain,oriTest,field,numField=[],postfix= postfix)
trainFile = oriTrain[0:-4]+ '_'+ postfix +'.ffm'
testFile = oriTest[0:-4]+ '_'+ postfix +'.ffm'
t=str(150)
k=str(8)
l=str(0.00002)
resultName = '../submit/ffm_'+ postfix + '_t{t}_k{k}_l{l}'.format(t=t,k=k,l=l,feat=str(len(ori11Field)))
assignCMDPara=' -t {t} -k {k} -l {l} -s 4 -r 0.02 '.format(t=t,k=k,l=l)
ffm.runffm(trainFile,testFile,resultName,assignCMDPara)
def run_userIDappID_2():
ori11Field=copyField[:]
oriTrain = '../data/temp/train_preAction_userID,appID_5_.csv'
oriTest = '../data/temp/test_preAction_userID,appID_5_.csv'
ori11Field.extend(['preAction','preClickCount','preClickDist','preConvCount','preConvDist'])
numField=['preClickCount','preClickDist','preConvCount','preConvDist']
postfix='m_p_u_amc_s{feat}_preAction_appIDuserID_5'.format(feat=str(len(ori11Field)))
csv_2_ffm.to_ffm(oriTrain,oriTest,ori11Field,numField=numField,postfix= postfix)
trainFile = oriTrain[0:-4]+ '_'+ postfix +'.ffm'
testFile = oriTest[0:-4]+ '_'+ postfix +'.ffm'
t=str(150) ;k=str(8) ;l=str(0.00002)
resultName = '../submit/ffm_'+ postfix + '_t{t}_k{k}_l{l}'.format(t=t,k=k,l=l,feat=str(len(ori11Field)))
assignCMDPara=' -t {t} -k {k} -l {l} -s 4 -r 0.02 '.format(t=t,k=k,l=l)
ffm.runffm(trainFile,testFile,resultName,assignCMDPara)
def joinXgbFmmRun(n):
trainFile = '../data/concat/xgb{n}_s11.ffm'.format(n=n)
testFile = '../data/concat/test_xgb{n}_s11.ffm'.format(n=n)
xgbTrain='../data/concat/xgb_trainFeature_FFMlibsvm_1_{n}'.format(n=n)
s11Train='../data/concat/m_p_u_amc_selected11.ffm'
#csv_2_ffm.joinOriXGB(xgbTrain,s11Train,trainFile)
xgbTest='../data/concat/xgb_testFeature_FFMlibsvm_1_{n}'.format(n=n)
s11Test='../data/concat/test_m_p_u_amc_selected11.ffm'
#csv_2_ffm.joinOriXGB(xgbTest,s11Test,testFile)
# t=str(150)
# k=str(8)
# l=str(0.00002)
# resultName = '../submit/ffm_joinXgb{n}-11_t{t}_k{k}_l{l}'.format(n=n,t=t,k=k,l=l)
# assignCMDPara=' -t {t} -k {k} -l {l} -s 4 -r 0.02 '.format(t=t,k=k,l=l)
# return ffm.runffm(trainFile,testFile,resultName,assignCMDPara)
#
t=str(200)
k=str(12)
l=str(0.000005)
resultName = '../submit/ffm_joinXgb{n}-11_t{t}_k{k}_l{l}'.format(n=n,t=t,k=k,l=l)
assignCMDPara=' -t {t} -k {k} -l {l} -s 3 -r 0.02 '.format(t=t,k=k,l=l)
return ffm.runffm(trainFile,testFile,resultName,assignCMDPara)
def combRun():
#--------------------55个组合特征 + 原始11特征使用--------------
field=combAllFeat()[0:55]
field.extend(ori11Field)
oriTrain = '../data/ori_and_11twoComb.csv'
oriTest = '../data/test_ori_and_11twoComb.csv'
postfix='11+55_s{feat}'.format(feat=str(len(field)))
# csv_2_ffm.to_ffm(oriTrain,oriTest,field,numField=[],postfix= postfix)
trainFile = oriTrain[0:-4]+ '_'+ postfix +'.ffm'
testFile = oriTest[0:-4]+ '_'+ postfix +'.ffm'
k=str(16)
l=str(0.000001)
#-200次迭代
t=str(200)
resultName = '../submit/ffmresult_'+ postfix + '_t{t}_k{k}_l{l}'.format(t=t,k=k,l=l,feat=str(len(ori11Field)))
assignCMDPara=' -t {t} -k {k} -l {l} -s 4 -r 0.02 '.format(t=t,k=k,l=l)
ffm.runffm(trainFile,testFile,resultName,assignCMDPara)
def mergeRun(field_21,
numField_21,
fixDay,
addPostfix='',
needMerge=True,
toffm=True):
# merge_train = merge_appID_userID('../data/temp/train_preAction_userID_{fixDay}.csv'.format(fixDay=fixDay),
# '../data/temp/train_preAction_userIDappID_{fixDay}.csv'.format(fixDay=fixDay),
# '../data/temp/merge_appID_userID_{fixDay}.csv'.format(fixDay=fixDay))
# merge_test = merge_appID_userID('../data/temp/test_preAction_userID_{fixDay}.csv'.format(fixDay=fixDay),
# '../data/temp/test_preAction_userIDappID_{fixDay}.csv'.format(fixDay=fixDay),
# '../data/temp/merge_test_appID_userID_{fixDay}.csv'.format(fixDay=fixDay))
merge_train = '../data/temp/old/merge_appID_userID_{fixDay}.csv'.format(
fixDay=fixDay)
merge_test = '../data/temp/old/merge_test_appID_userID_{fixDay}.csv'.format(
fixDay=fixDay)
postfix = 'mergeAppUser_s{feat}_preAction_{fixDay}_{addPostfix}'.format(
feat=str(len(field_21)), fixDay=fixDay, addPostfix=addPostfix)
toffm = False
if toffm:
csv_2_ffm.to_ffm(merge_train,
merge_test,
field_21,
numField=numField_21,
postfix=postfix)
trainFile = merge_train[0:-4] + '_' + postfix + '.ffm'
testFile = merge_test[0:-4] + '_' + postfix + '.ffm'
t = str(120)
k = str(8)
l = str(0.00002)
resultName = '../submit/ffm_' + postfix + '_t{t}_k{k}_l{l}'.format(
t=t, k=k, l=l, feat=str(len(field_21)))
assignCMDPara = ' -t {t} -k {k} -l {l} -s 4 -r 0.02 '.format(t=t, k=k, l=l)
ffm.runffm(trainFile, testFile, resultName, assignCMDPara)
def oriInRun():
oriTrain = '../data/sample/recent20.csv'
oriTest = '../data/sample/test_oldPositionIDadID.csv'
postfix='oldPositionIDadID_recent20_s{feat}'.format(feat=str(len(ori11Field)))
csv_2_ffm.to_ffm(oriTrain,oriTest,ori11Field,numField=[],postfix= postfix)
trainFile = oriTrain[0:-4]+ '_'+ postfix +'.ffm'
testFile = oriTest[0:-4]+ '_'+ postfix +'.ffm'
t=str(150)
k=str(8)
l=str(0.00002)
resultName ='../data/sample/result/ffm_'+ postfix + '_t{t}_k{k}_l{l}'.format(t=t,k=k,l=l,feat=str(len(ori11Field)))
assignCMDPara=' -t {t} -k {k} -l {l} -s 4 -r 0.02 '.format(t=t,k=k,l=l)
instanceID=pd.read_csv(oriTest)['instanceID'].tolist()
return ffm.runffm(trainFile,testFile,resultName,assignCMDPara,instanceID=instanceID)
def newInRun():
newInField=ori11Field[:]
newInField.extend( [ "hometown" ,"residence","sitesetID","appPlatform" "hourTime","appSum"] )
oriTrain = '../data/sample/mergeAll_recent20.csv'
oriTest = '../data/sample/test_newPositionIDadID.csv'
postfix='newPositionIDadID_recent20_s{feat}'.format(feat=str(len(newInField)))
csv_2_ffm.to_ffm(oriTrain,oriTest,ori11Field,numField=[],postfix= postfix)
trainFile = oriTrain[0:-4]+ '_'+ postfix +'.ffm'
testFile = oriTest[0:-4]+ '_'+ postfix +'.ffm'
t=str(150)
k=str(8)
l=str(0.00002)
resultName = '../data/sample/result/ffm_'+ postfix + '_t{t}_k{k}_l{l}'.format(t=t,k=k,l=l,feat=str(len(ori11Field)))
assignCMDPara=' -t {t} -k {k} -l {l} -s 4 -r 0.02 '.format(t=t,k=k,l=l)
instanceID=pd.read_csv(oriTest)['instanceID'].tolist()
return ffm.runffm(trainFile,testFile,resultName,assignCMDPara,instanceID=instanceID)
'connectionType_appID', #44
'adID_advertiserID', #45
'positionType_advertiserID', #46
'camgaignID_appID', #47
'positionType_appCategory', #48
'advertiserID_appID', #49
'positionType_appID', #50
'adID_appID', #51
'camgaignID_appCategory', #52
'appID_appCategory', #53
'adID_appCategory', #54
'advertiserID_appCategory', #55
]
oriTrain = '../data/m_p_u_amc_only11twoComb.csv'
oriTest = '../data/test_m_p_u_amc_only11twoComb.csv'
featNum = 22
field = combAllFeat()[0:featNum]
csv_2_ffm.to_ffm(oriTrain, oriTest, field, numField=[], postfix=str(featNum))
trainFile = '../data/m_p_u_amc_only11twoComb' + '_' + str(featNum) + '.ffm'
testFile = '../data/test_m_p_u_amc_only11twoComb' + '_' + str(featNum) + '.ffm'
t = str(250)
resultName = '../submit/ffmresult_ori_and_11twoComb' + '_' + str(
featNum) + '_t' + t
assignCMDPara = './ffm-train -l 0.00002 -k 4 -t ' + t + ' -r 0.02 -s 8'
ffm.runffm(trainFile, testFile, resultName, assignCMDPara)
#--------------------------