def __mkMap(key, nf, ni, ef1, ef2, ei, oFile):
# leaf nodeの構築
infL = [
nm.mcommon(k=ef1, K=key, m=ei, r=True, i=ei).mcut(f="%s:nam" % (ef1)),
nm.mcommon(k=ef2, K=key, m=ei, r=True, i=ei).mcut(f="%s:nam" % (ef2))
]
if ni:
infL.append(
nm.mcommon(k=nf, K=key, m=ei, r=True,
i=ni).mcut(f="%s:nam" % (nf)))
xleaf = nm.muniq(i=infL, k="nam")
xleaf <<= nm.msetstr(v=1, a="leaf")
if ni:
inp = [
nm.mcut(f="%s:nam" % (nf), i=ni),
nm.mcut(f="%s:nam" % (key), i=ni)
]
else:
inp = [
nm.mcut(f="%s:nam" % (ef1), i=ei),
nm.mcut(f="%s:nam" % (ef2), i=ei),
nm.mcut(f="%s:nam" % (key), i=ei)
]
f = None
f <<= nm.muniq(k="nam", i=inp)
f <<= nm.mjoin(k="nam", m=xleaf, f="leaf", n=True)
# nullは最初に来るはずなので、mcalでなくmnumberでもnullを0に採番できるはずだが念のために
f <<= nm.mcal(c='if(isnull($s{nam}),0,line()+1)', a="num")
f <<= nm.mnullto(f="nam", v="##NULL##", o=oFile)
f.run()
def mnest2tree(ei, ef, k, ni=None, nf=None, ev=None, no=None, eo=None):
# paracheck追加
efs = ef.split(",")
ef1 = efs[0]
ef2 = efs[1]
f = nm.mcut(f="%s:#orgKey,%s:#orgEf1,%s:#orgEf2" % (k, ef1, ef2), i=ei)
temp = mtemp.Mtemp()
of = temp.file()
with _nu.mcsvout(o=of, f="#orgKey,#orgEf1,#orgEf2,#ef1,#ef2") as oCSV:
for flds in f:
orgKey = flds[0]
orgEf1 = flds[1]
orgEf2 = flds[2]
oCSV.write([orgKey, orgEf1, orgEf2, orgKey, orgEf1])
oCSV.write([orgKey, orgEf1, orgEf2, orgKey, orgEf2])
f = None
f <<= nm.mjoin(k="#orgKey,#orgEf1,#orgEf2", K=[k, ef1, ef2], m=ei,
i=of) # 全項目join
if ev:
f <<= nm.mavg(k="#ef1,#ef2", f=ev)
else:
f <<= nm.muniq(k="#ef1,#ef2")
f <<= nm.mcut(r=True, f="#orgKey,#orgEf1,#orgEf2")
f <<= nm.mfldname(f="#ef1:%s,#ef2:%s" % (ef1, ef2), o=eo)
f.run()
if ni:
head = nu.mheader(i=ni)
fldnames = [s for s in head if s != nf]
commas = ',' * (len(fldnames) - 1)
f0 = None
f0 <<= nm.mcut(f="%s:%s" % (ef1, nf), i=eo)
f0 <<= nm.muniq(k=nf)
f0 <<= nm.mcommon(k=nf, m=ni, r=True)
f0 <<= nm.msetstr(v=commas, a=fldnames)
f = nm.mcut(f=k, r=True, i=[ni, f0])
f <<= nm.msetstr(v="", a=k, o=no)
f.run()
def run(self, **kw_args):
os.environ["KG_VerboseLevel"] = "2"
if "msg" in kw_args:
if kw_args["msg"] == "on":
os.environ['KG_ScpVerboseLevel'] = "3"
temp = Mtemp()
xxedge = temp.file()
xxnode = temp.file()
xxnam2num = temp.file()
xxnum2nam = temp.file()
xxebase = temp.file()
xxbody = temp.file()
e1 = None
if self.ew:
e1 <<= nm.mcut(f="%s:__node1,%s:__node2,%s:__weight" %
(self.ef1, self.ef2, self.ew),
i=self.eFile)
else:
e1 <<= nm.mcut(f="%s:__node1,%s:__node2" % (self.ef1, self.ef2),
i=self.eFile)
e1 <<= nm.muniq(k="__node1,__node2")
e2 = nm.mfldname(i=e1, f="__node2:__node1,__node1:__node2")
fe = None
fe <<= nm.muniq(k="__node1,__node2", i=[e1, e2], o=xxedge)
fe.run()
# cleaning the node data (remove duplicate nodes)
fn = None
if self.nFile:
if self.nw:
fn <<= nm.mcut(f="%s:__node,%s" % (self.nf, self.nw),
i=self.nFile)
else:
fn <<= nm.mcut(f="%s:__node" % (self.nf), i=self.nFile)
fn <<= nm.muniq(k="__node", o=xxnode)
else:
xxen1 = nm.mcut(f="__node1:__node", i=xxedge)
xxen2 = nm.mcut(f="__node2:__node", i=xxedge)
fn <<= nm.muniq(k="__node", o=xxnode, i=[xxen1, xxen2])
fn.run()
# 節点名<=>節点番号変換表の作成
fmap = None
fmap <<= nm.mcut(f="__node", i=xxnode)
fmap <<= nm.mnumber(a="__num", S=1, q=True, o=xxnam2num)
fmap <<= nm.msortf(f="__num", o=xxnum2nam)
fmap.run()
# 節点ファイルが指定された場合は枝ファイルとの整合性チェック
if self.nFile:
ncheck = nm.mcut(f="__node1:__node", i=xxedge)
ncheck <<= nm.mcommon(k="__node", m=xxnam2num, r=True)
nmatch = ncheck.run()
if len(nmatch) > 0:
raise Exception(
"#ERROR# the node named '%s' in the edge file doesn't exist in the node file."
% (nmatch[0][0]))
# metisのグラフファイルフォーマット
# 先頭行n m [fmt] [ncon]
# n: 節点数、m:枝数、ncon: 節点weightの数
# 1xx: 節点サイズ有り (not used, meaning always "0")
# x1x: 節点weight有り
# xx1: 枝がweightを有り
# s w_1 w_2 ... w_ncon v_1 e_1 v_2 e_2 ... v_k e_k
# s: 節点サイズ (節点サイズは利用不可)
# w_x: 節点weight
# v_x: 接続のある節点番号(行番号)
# e_x: 枝weight
# --------------------
# generate edge data using the integer numbered nodes
#fnnum = None
fnnum = nm.mcut(f="__num:__node_n1", i=xxnam2num) # {xxnnum}
fenum = None
fenum <<= nm.mjoin(k="__node1",
K="__node",
f="__num:__node_n1",
m=xxnam2num,
i=xxedge)
fenum <<= nm.mjoin(k="__node2",
K="__node",
f="__num:__node_n2",
m=xxnam2num)
fenum <<= nm.msortf(f="__node_n1") #{xxenum}
febase = None
febase <<= nm.mnjoin(k="__node_n1", m=fenum, i=fnnum, n=True)
febase <<= nm.msortf(f="__node_n1%n,__node_n2%n",
o=xxebase) #{xxebase}"
febase.run()
fbody = None
if not self.ew:
fbody <<= nm.mcut(f="__node_n1,__node_n2", i=xxebase)
fbody <<= nm.mtra(k="__node_n1", f="__node_n2", q=True)
fbody <<= nm.mcut(f="__node_n2", nfno=True, o=xxbody)
# if ew= is specified, merge the weight data into the edge data.
else:
febody = None
febody <<= nm.mcut(f="__node_n1,__node_n2:__v", i=xxebase)
febody <<= nm.mnumber(S=0, I=2, a="__seq", q=True)
fwbody = None
fwbody <<= nm.mcut(f="__node_n1,__weight:__v", i=xxebase)
fwbody <<= nm.mnumber(S=1, I=2, a="__seq", q=True)
fbody <<= nm.msortf(f="__seq%n", i=[febody, fwbody])
fbody <<= nm.mtra(k="__node_n1", f="__v", q=True)
fbody <<= nm.mcut(f="__v", nfno=True, o=xxbody)
fbody.run()
# xxbody
# 2 7 3 8 5 9
# 1 7 3 10 5 11 7 12
# 1 8 2 10 4 13 7 14
# --------------------
# generate node data using integer number
if self.nFile and self.nw:
# xxnode
# __node,v1,v2
# a,1,1
# b,1,1
# c,1,1
xxnbody = temp.file()
xxnbody1 = temp.file()
fnbody = None
fnbody <<= nm.mjoin(k="__node", f="__num", i=xxnode, m=xxnam2num)
fnbody <<= nm.msortf(f="__num%n")
fnbody <<= nm.mcut(f=self.nw, nfno=True)
fnbody <<= nm.cmd("tr ',' ' ' ") # tricky!!
fnbody <<= nm.mwrite(o=xxnbody)
fnbody.run()
# xxnbody
# 1 1
# 1 1
# 1 1
# paste the node weight with edge body
fnbody1 = None
fnbody1 <<= nm.mpaste(nfn=True, m=xxbody, i=xxnbody)
fnbody1 <<= nm.cmd("tr ',' ' ' ")
fnbody1 <<= nm.mwrite(o=xxnbody1)
fnbody1.run()
os.system("mv %s %s" % (xxnbody1, xxbody))
# xxbody
# 1 1 2 7 3 8 5 9
# 1 1 1 7 3 10 5 11 7 12
# 1 1 1 8 2 10 4 13 7 14
eSize = mrecount(i=xxedge)
eSize /= 2
nSize = mrecount(i=xxnode)
nwFlag = 1 if self.nw else 0
ewFlag = 1 if self.ew else 0
fmt = "0%d%d" % (nwFlag, ewFlag)
xxhead = temp.file()
xxgraph = temp.file()
os.system("echo '%d %d %s %d' > %s" %
(nSize, eSize, fmt, self.ncon, xxhead))
os.system("cat %s %s > %s" % (xxhead, xxbody, xxgraph))
if self.mFile:
nm.mfldname(f="__num:num,__node:node", i=xxnum2nam,
o=self.mFile).run()
if self.dFile:
os.system("cp %s %s" % (xxgraph, self.dFile))
if not self.noexe:
if self.verbose:
os.system(
"gpmetis -seed=%d -ptype=%s -ncuts=%d -ufactor=%d %s %d" %
(self.seed, self.ptype, self.ncuts, self.ufactor, xxgraph,
self.kway))
else:
os.system(
"gpmetis -seed=%d -ptype=%s -ncuts=%d -ufactor=%d %s %d > /dev/null"
% (self.seed, self.ptype, self.ncuts, self.ufactor,
xxgraph, self.kway))
import glob
if len(glob.glob(xxgraph + ".part.*")) == 0:
raise Exception(
"#ERROR# command `gpmetis' didn't output any results")
# 節点名を数字から元に戻す
# #{xxgraph}.part.#{kway}
# 1
# 0
# 1
fo = None
fo <<= nm.mcut(f="0:cluster",
nfni=True,
i=xxgraph + ".part." + str(self.kway))
fo <<= nm.mnumber(S=1, a="__num", q=True)
fo <<= nm.mjoin(k="__num", f="__node", m=xxnum2nam)
fo <<= nm.msortf(f="__node,cluster")
if self.nf:
fo <<= nm.mcut(f="__node:%s,cluster" % (self.nf), o=self.oFile)
else:
fo <<= nm.mcut(f="__node:node,cluster", o=self.oFile)
fo.run()
nu.mmsg.endLog(self.__cmdline())
tra <<= nm.mcut(f="id,date2:date,ret")
# frequency of one item
freq = None
freq <<= nm.mcut(f="id", i=tra)
freq <<= nm.mcount(k="id", a="freq")
freq <<= nm.mselnum(f="freq", c="[5,]")
# total number of transactions
total = None
total <<= nm.mcut(f="date", i=tra)
total <<= nm.muniq(k="date")
total <<= nm.mcount(a="total")
# frequency of cooccurence of id, and calculate lift values
itemCoFreq = None
itemCoFreq <<= nm.mcut(f="date,id", i=tra)
itemCoFreq <<= nm.mcommon(k="id", m=freq)
itemCoFreq <<= nm.mcombi(k="date", n=2, f="id", a="id1,id2")
itemCoFreq <<= nm.mcut(f="id1,id2")
itemCoFreq <<= nm.mfsort(f="id1,id2")
itemCoFreq <<= nm.mcount(k="id1,id2", a="coFreq")
itemCoFreq <<= nm.mjoin(k="id1", m=freq, K="id", f="freq:freq1")
itemCoFreq <<= nm.mjoin(k="id2", m=freq, K="id", f="freq:freq2")
itemCoFreq <<= nm.mproduct(m=total, f="total")
itemCoFreq <<= nm.mcal(c="(${coFreq}*${total})/(${freq1}*${freq2})", a="lift")
itemCoFreq <<= nm.msel(c="${lift}>10 && ${coFreq}>6")
r = itemCoFreq.run(msg=debug)
print(r)
print(len(r))
def enumerate(self,eArgs):
tf=nu.Mtemp()
# 最小サポートと最小サポート件数
if "minCnt" in eArgs :
self.minCnt = int(eArgs["minCnt"])
self.minSup = float(self.minCnt)/ float(self.db.size)
else:
self.minSup = float(eArgs["minSup"])
self.minCnt = int(self.minSup * float(self.db.size) + 0.99)
# 最大サポートと最大サポート件数
self.maxCnt=None
if "maxCnt" in eArgs or "maxSup" in eArgs:
if "maxCnt" in eArgs:
self.maxCnt = int(eArgs["maxCnt"])
self.maxSup = float(self.maxCnt)/float(self.db.size)
else:
self.maxSup = float(eArgs["maxSup"])
self.maxCnt = int(self.maxSup * float(self.db.size) + 0.99)
#未使用
#@minProb = eArgs["minProb"].to_f # 事後確率
#@minGR = @minProb/(1-@minProb) # 増加率
#@minGR = eArgs["minGR"].to_f if eArgs["minGR"]
# あるクラスをpos、他のクラスをnegにして、パターン列挙した結果ファイル名を格納する
pFiles=[]
tFiles=[]
for cName,posSize in self.db.clsNameRecSize.items():
negSize=self.db.size-posSize
# minGRの計算
if "minGR" in eArgs:
self.minGR=eArgs["minGR"]
else:
minProb = eArgs["minProb"] if ( "minProb" in eArgs ) else 0.5
if "uniform" in eArgs and eArgs["uniform"]:
self.minGR = (minProb/(1-minProb)) * (self.db.clsSize-1) # マニュアルの式(4)
else:
self.minGR = (minProb/(1-minProb)) * (float(negSize)/float(posSize)) # マニュアルの式(4)
# 最小サポートと最小サポート件数
if "minCnt" in eArgs:
self.minPos = eArgs["minCnt"]
else:
self.minPos = int(eArgs["minSup"] * float(posSize) + 0.99)
# 最大サポートと最大サポート件数
if "maxCnt" in eArgs or "maxSup" in eArgs:
if "maxCnt" in eArgs:
self.maxCnt = int(eArgs["maxCnt"])
else:
self.maxCnt = int(eArgs["maxSup"] * float(posSize) + 0.99)
self.sigma[cName] = self.calSigma(self.minPos,self.minGR,posSize,negSize)
# lcm_seqのパラメータ設定と実行
lcmout = tf.file() # lcm_seq出力ファイル
# 頻出パターンがなかった場合、lcm出力ファイルが生成されないので
# そのときのために空ファイルを生成しておいく。
with open(lcmout, "w") as efile:
pass
params = {}
if self.msgoff:
params["type"] ="CIA_"
else:
params["type"] ="CIA"
if self.maxCnt: # windowサイズ上限
params["U"] = str(self.maxCnt)
if "minLen" in eArgs:
params["l"] = str(eArgs["minLen"])
if 'maxLen' in eArgs:
params["u"] = str(eArgs["maxLen"])
if 'gap' in eArgs:
params["g"] = str(eArgs["gap"])
if 'win' in eArgs:
params["G"] = str(eArgs["win"])
params["w"] = self.weightFile[cName]
params["i"] = self.file
params["sup"] = str(self.sigma[cName])
params["o"] = lcmout
# lcm_seq実行
#MCMD::msgLog("#{run}")
if 'padding' in eArgs and eArgs["padding"]: # padding指定時は、0アイテムを出力しないlcm_seqを実行
extTake.lcmseq_zero(params)
else:
extTake.lcmseq(params)
# パターンのサポートを計算しCSV出力する
#MCMD::msgLog("output patterns to CSV file ...")
pFiles.append(self.temp.file())
transle = self.temp.file()
extTake.lcmtrans(lcmout,"e",transle) # pattern,countP,countN,size,pid
f=None
f <<= nm.mdelnull(f="pattern",i=transle)
f <<= nm.mcal(c='round(${countN},1)',a="neg")
f <<= nm.mcal(c='round(${countP}/%s,1)'%(self.posWeight[cName]),a="pos")
f <<= nm.mdelnull(f="pattern")
f <<= nm.msetstr(v=cName,a="class")
f <<= nm.msetstr(v=posSize,a="posTotal")
f <<= nm.msetstr(v=self.minGR,a="minGR")
f <<= nm.mcut(f="class,pid,pattern,size,pos,neg,posTotal,minGR",o=pFiles[-1])
f.run()
#s = MCMD::mrecount("i=#{pFiles.last}") # 列挙されたパターンの数
#MCMD::msgLog("the number of contrast patterns on class `#{cName}' enumerated is #{s}")
if self.outtf :
# トランザクション毎に出現するシーケンスを書き出す
#MCMD::msgLog("output tid-patterns ...")
tFiles.append(self.temp.file())
xxw= tf.file()
f=None
f <<= nm.mcut(f=self.db.idFN,i=self.db.file)
f <<= nm.muniq(k=self.db.idFN)
f <<= nm.mnumber(S=0,a="__tid",q=True)
f <<= nm.msortf(f="__tid",o=xxw)
f.run()
nm.mcut(f=self.db.idFN,i=self.db.file).muniq(k=self.db.idFN).mnumber(S=0,a="__tid",q=True,o=xxw).run()
translt = self.temp.file()
extTake.lcmtrans(lcmout,"t",translt)
nm.mjoin(k="__tid",m=xxw,f=self.db.idFN,i=translt).msetstr(v=cName,a="class").mcut(f=self.db.idFN+",class,pid",o=tFiles[-1]).run()
# クラス別のパターンとtid-pidファイルを統合して最終出力
self.pFile = self.temp.file()
self.tFile = self.temp.file()
# パターンファイル併合
xxpCat = tf.file()
f = nm.mcat(i=",".join(pFiles))
f <<= nm.msortf(f="class,pid")
f <<= nm.mnumber(s="class,pid",S=0,a="ppid",o=xxpCat)
f.run()
# パターンファイル計算
items=self.db.items
f=""
f = nm.mcut(f="class,ppid:pid,pattern,size,pos,neg,posTotal,minGR",i=xxpCat)
f <<= nm.msetstr(v=self.db.size,a="total")
f <<= nm.mcal(c='${total}-${posTotal}',a="negTotal") # negのトータル件数
f <<= nm.mcal(c='${pos}/${posTotal}',a="support") # サポートの計算
f <<= nm.mcal(c='if(${neg}==0,1.797693135e+308,(${pos}/${posTotal})/(${neg}/${negTotal}))',a="growthRate")
if "uniform" in eArgs and eArgs["uniform"] == True:
f <<= nm.mcal(c='(${pos}/${posTotal})/(${pos}/${posTotal}+(%s-1)*${neg}/${negTotal})'%(self.db.clsSize),a="postProb")
else:
f <<= nm.mcal(c='${pos}/(${pos}+${neg})',a="postProb")
f <<= nm.msel(c='${pos}>=%s&&${growthRate}>=${minGR}'%(self.minPos)) # minSupとminGRによる選択
f <<= nm.mvreplace(vf="pattern",m=items.file,K=items.idFN,f=items.itemFN)
f <<= nm.mcut(f="class,pid,pattern,size,pos,neg,posTotal,negTotal,total,support,growthRate,postProb")
f <<= nm.mvsort(vf="pattern")
f <<= nm.msortf(f="class%nr,postProb%nr,pos%nr",o=self.pFile)
f.run()
if self.outtf :
# 列挙されたパターンを含むtraのみ選択するためのマスタ
xxp4=nm.mcut(f="class,pid",i=self.pFile)
f = nm.mcat(i=",".join(tFiles))
f <<= nm.mjoin(k="class,pid",m=xxpCat,f="ppid") # 全クラス統一pid(ppid)結合
f <<= nm.mcommon(k="class,ppid",K="class,pid",m=xxp4) # 列挙されたパターンの選択
f <<= nm.mcut(f=self.db.idFN+",class,ppid:pid")
f <<= nm.msortf(f=self.db.idFN+",class,pid",o=self.tFile)
f.run()
self.size = nu.mrecount(i=self.pFile)
def run(self):
from datetime import datetime
t = datetime.now()
wf = nu.Mtemp()
xxinp = wf.file()
xxmap = wf.file()
input = self.ei
self.g2pair(self.ni, self.nf, self.ei, self.ef1, self.ef2, xxinp,
xxmap)
xxmace = wf.file() # maceの出力(tra形式)
para = {}
if self.msgoff:
para["type"] = "Ce_" if self.all else "Me_"
else:
para["type"] = "Ce" if self.all else "Me"
para["i"] = xxinp
para["o"] = xxmace
if self.minSize:
para["l"] = self.minSize
if self.maxSize:
para["u"] = self.maxSize
extTake.mace(para)
#MCMD::msgLog("converting the numbered nodes into original name ...")
id = nu.mrecount(i=xxmace, nfni=True)
# xxpair = wf.file() # 上記traをpair形式に変換したデータ
fpair = None
fpair <<= nm.mcut(i=xxmace, nfni=True, f="0:num")
fpair <<= nm.mnumber(q=True, a="id")
fpair <<= nm.mvcount(vf="num:size")
fpair <<= nm.mtra(r=True, f="num")
# when ni= specified, it add the isolated single cliques.
if self.ni:
fpair_u = nm.mread(i=fpair)
if self.all:
fpair_u <<= nm.mselstr(f="size", v=1)
fpair_u <<= nm.mcut(f="num")
fpair_u <<= nm.muniq(k="num")
# select all nodes which are not included in any cliques
xxiso = None
xxiso <<= nm.mcut(f="num", i=xxmap)
xxiso <<= nm.mcommon(k="num", m=fpair_u, r=True)
xxiso <<= nm.mnumber(S=id, a="id", q=True)
xxiso <<= nm.msetstr(v=1, a="size")
xxiso <<= nm.mcut(f="id,num,size")
# mcut入れないとおかしくなるあとで直す
#ddlist = [fpair.mcut(f="id,num,size"),xxiso]
xxpair = nm.mcut(i=[fpair, xxiso], f="id,num,size")
else:
xxpair = fpair
xxpair <<= nm.mjoin(m=xxmap, k="num", f="node")
xxpair <<= nm.mcut(f="id,node,size")
xxpair <<= nm.msortf(f="id,node", o=self.oFile)
xxpair.run()
procTime = datetime.now() - t
# ログファイル出力
if self.logFile:
kv = [["key", "value"]]
for k, v in self.args.items():
kv.append([k, str(v)])
kv.append(["time", str(procTime)])
nm.writecsv(i=kv, o=self.logFile).run()
def enumerate(self, eArgs):
pFiles = []
tFiles = []
tf = mtemp.Mtemp()
for cName, posSize in self.db.clsNameRecSize.items():
negSize = self.db.traSize - posSize
if "minGR" in eArgs:
self.minGR = eArgs["minGR"]
else:
minProb = eArgs["minProb"] if ("minProb" in eArgs) else 0.5
if "uniform" in eArgs and eArgs["uniform"] == True:
self.minGR = (minProb / (1 - minProb)) * (
self.db.clsSize - 1) # マニュアルの式(4)
else:
self.minGR = (minProb / (1 - minProb)) * (
float(negSize) / float(posSize)) # マニュアルの式(4)
# 最小サポートと最小サポート件数
# s=0.05
# s=c1:0.05,c2:0.06
# S=10
# S=c1:10,c2:15
if "minCnt" in eArgs:
if isinstance(eArgs["minCnt"], dict):
self.minPos = eArgs["minCnt"][cName]
else:
self.minPos = eArgs["minCnt"]
else:
if isinstance(eArgs["minSup"], dict):
self.minPos = int(eArgs["minSup"][cName] * float(posSize) +
0.99)
else:
self.minPos = int(eArgs["minSup"] * flost(posSize) + 0.99)
# 最大サポートと最大サポート件数
if "maxCnt" in eArgs:
if isinstance(eArgs["maxCnt"], dict):
self.maxPos = eArgs["maxCnt"][cName]
else:
self.maxPos = eArgs["maxCnt"]
elif "maxSup" in eArgs:
if isinstance(eArgs["maxSup"], dict):
self.maxPos = int(eArgs["maxSup"][cName] * float(posSize) +
0.99)
else:
self.maxPos = int(eArgs["maxSup"] * float(posSize) + 0.99)
else:
self.maxPos = None
self.sigma[cName] = self.calSigma(self.minPos, self.minGR, posSize,
negSize)
# lcmのパラメータ設定と実行
# 頻出パターンがなかった場合、lcm出力ファイルが生成されないので
# そのときのために空ファイルを生成しておいく。
lcmout = tf.file() # lcm出力ファイル
with open(lcmout, "w") as efile:
pass
runPara = {}
if self.msgoff:
runPara["type"] = eArgs["type"] + "IA_"
else:
runPara["type"] = eArgs["type"] + "IA"
#if self.maxPos: #rubyだとif @maxCntなってる（どこにも設定されてないので）動いてないはず
if self.maxPos:
runPara["U"] = self.maxPos
if "minLen" in eArgs:
runPara["l"] = str(eArgs["minLen"])
if "maxLen" in eArgs:
runPara["u"] = str(eArgs["maxLen"])
runPara["w"] = self.weightFile[cName]
runPara["i"] = self.file
runPara["sup"] = str(self.sigma[cName])
runPara["o"] = lcmout
# lcm実行
#MCMD::msgLog("#{run}")
#TAKE::run_lcm(run)
#print(self.sigma)
#print(runPara)
#MCMD::msgLog("output patterns to CSV file ...")
extTake.lcm(runPara)
pFiles.append(self.temp.file())
transle = tf.file()
extTake.lcmtrans(lcmout, "e", transle)
f = nm.mdelnull(f="pattern", i=transle)
f <<= nm.mcal(c='round(${countN},1)', a="neg")
f <<= nm.mcal(c='round(${countP}/%s,1)' % (self.posWeight[cName]),
a="pos")
f <<= nm.mdelnull(f="pattern") #いる？
f <<= nm.msetstr(v=cName, a="class")
f <<= nm.msetstr(v=posSize, a="posTotal")
f <<= nm.msetstr(v=self.minGR, a="minGR")
f <<= nm.mcut(f="class,pid,pattern,size,pos,neg,posTotal,minGR",
o=pFiles[-1])
f.run()
#s = nutil.mrecount(i=self.file)
#MCMD::msgLog("the number of contrast patterns on class `#{cName}' enumerated is #{s}")
if self.outtf:
# トランザクション毎に出現するパターンを書き出す
#MCMD::msgLog("output tid-patterns ...")
tFiles.append(self.temp.file())
xxw = tf.file()
xxw = nm.mcut(f=self.db.idFN, i=self.db.file)
xxw <<= nm.muniq(k=self.db.idFN)
xxw <<= nm.mnumber(S=0, a="__tid", q=True)
translt = self.temp.file()
extTake.lcmtrans(lcmout, "t", translt)
f = nm.mjoin(k="__tid", m=xxw, f=self.db.idFN, i=translt)
f <<= nm.msetstr(v=cName, a="class")
f <<= nm.mcut(f=self.db.idFN + ",class,pid", o=tFiles[-1])
f.run()
# クラス別のパターンとtid-pidファイルを統合して最終出力
self.pFile = self.temp.file()
self.tFile = self.temp.file()
# パターンファイル併合
xxpCat = tf.file()
f = nm.mcat(i=",".join(pFiles))
f <<= nm.msortf(f="class,pid")
f <<= nm.mnumber(s="class,pid", S=0, a="ppid", o=xxpCat)
f.run()
# パターンファイル計算
items = self.db.items
f = nm.mcut(f="class,ppid:pid,pattern,size,pos,neg,posTotal,minGR",
i=xxpCat)
f <<= nm.msetstr(v=self.db.traSize, a="total")
f <<= nm.mcal(c='${total}-${posTotal}', a="negTotal") # negのトータル件数
f <<= nm.mcal(c='${pos}/${posTotal}', a="support") # サポートの計算
f <<= nm.mcal(
c=
'if(${neg}==0,1.797693135e+308,(${pos}/${posTotal})/(${neg}/${negTotal}))',
a="growthRate")
if "uniform" in eArgs and eArgs["uniform"] == True:
f <<= nm.mcal(
c='(${pos}/${posTotal})/(${pos}/${posTotal}+(%s-1)*${neg}/${negTotal})'
% (self.db.clsSize),
a="postProb")
else:
f <<= nm.mcal(c='${pos}/(${pos}+${neg})', a="postProb")
f <<= nm.msel(c='${pos}>=%s&&${growthRate}>=${minGR}' %
(self.minPos)) # minSupとminGRによる選択
f <<= nm.mvreplace(vf="pattern",
m=items.file,
K=items.idFN,
f=items.itemFN)
f <<= nm.mcut(
f="class,pid,pattern,size,pos,neg,posTotal,negTotal,total,support,growthRate,postProb"
)
f <<= nm.mvsort(vf="pattern")
f <<= nm.msortf(f="class%nr,postProb%nr,pos%nr", o=self.pFile)
f.run()
# アイテムを包含している冗長なタクソノミを削除
if items.taxonomy:
#MCMD::msgLog("reducing redundant rules in terms of taxonomy ...")
##ここは後で
zdd = VSOP.constant(0)
dt = nm.mcut(i=self.pFile, f="pattern")
for fldVal in dt:
zdd = zdd + VSOP.itemset(fldVal[0])
zdd = self.reduceTaxo(zdd, self.db.items)
xxp1 = tf.file()
xxp2 = tf.file()
xxp3 = tf.file()
zdd.csvout(xxp1)
nm.mcut(nfni=True, f="1:pattern",
i=xxp1).mvsort(vf="pattern").msortf(f="pattern",
o=xxp2).run()
nm.msortf(f="pattern", i=self.pFile).mcommon(
k="pattern", m=xxp2).msortf(f="class%nr,postProb%nr,pos%nr",
o=xxp3).run()
shutil.move(xxp3, self.pFile)
if self.outtf:
# 列挙されたパターンを含むtraのみ選択するためのマスタ
xxp4 = nm.mcut(f="class,pid", i=self.pFile)
f = nm.mcat(i=",".join(tFiles))
f <<= nm.mjoin(k="class,pid", m=xxpCat,
f="ppid") # 全クラス統一pid(ppid)結合
f <<= nm.mcommon(k="class,ppid", K="class,pid",
m=xxp4) # 列挙されたパターンの選択
f <<= nm.mcut(f=self.db.idFN + ",class,ppid:pid")
f <<= nm.msortf(f=self.db.idFN + ",class,pid", o=self.tFile)
f.run()
def run(self):
temp = mtemp.Mtemp()
### mtra2gc
xxsimgN = temp.file()
xxsimgE = temp.file()
xxsimgE0 = temp.file()
param = {}
param["i"] = self.iFile
if self.idFN:
param["tid"] = self.idFN
if self.itemFN:
param["item"] = self.itemFN
if self.sp1:
param["s"] = self.sp1
if self.sp2:
param["S"] = self.sp2
#####################
# 異なる向きのconfidenceを列挙するためにsim=C th=0として双方向列挙しておく
# 出力データは倍になるが、mfriendsで-directedとすることで元が取れている
param["sim"] = "C"
param["th"] = "0"
param["node_support"] = True
if self.numtp:
param["num"] = True
param["no"] = xxsimgN
param["eo"] = xxsimgE0
nt.mtra2gc(**param).run()
f = nm.readcsv(xxsimgE0)
for i in range(self.filterSize):
f <<= nm.mselnum(f=self.filter[i],
c="[%s,%s]" % (self.lb[i], self.ub[i]))
f <<= nm.writecsv(xxsimgE)
f.run()
### mfrirends
xxfriends = temp.file()
xxfriendE = temp.file()
xxw = temp.file()
xxf = temp.file()
xxff = temp.file()
xxor = temp.file()
if not os.path.isdir(xxfriends):
os.makedirs(xxfriends)
col = [["FF000080", "FF888880"], ["0000FF80", "8888FF80"],
["00FF0080", "88FF8880"]]
for i in range(len(self.sim)):
paramf = {}
paramf["ei"] = xxsimgE
paramf["ni"] = xxsimgN
paramf["ef"] = "node1,node2"
paramf["nf"] = "node"
paramf["eo"] = xxfriendE
paramf["no"] = xxfriends + "/n_" + str(i)
paramf["sim"] = self.sim[i]
paramf["dir"] = self.dir[i]
paramf["rank"] = self.rank[i]
paramf["directed"] = True
nt.mfriends(**paramf).run()
frec2 = nm.mfsort(f="node1,node2", i=xxfriendE)
frec2 <<= nm.msummary(k="node1,node2",
f=self.sim[i],
c="count,mean")
frec2 <<= nm.mselstr(f="count", v=2)
# node1%0,node2%1,fld,count,mean
# a,b,support,2,0.1818181818
# a,d,support,2,0.1818181818
f = nm.mjoin(k="node1,node2",
K="node1,node2",
m=frec2,
f="mean:s1",
n=True,
i=xxfriendE)
f <<= nm.mjoin(k="node2,node1",
K="node1,node2",
m=frec2,
f="mean:s2",
n=True)
# 1) xxrecs2でsimをjoinできない(s1,s2共にnull)ということは、それは片方向枝なので"F"をつける
# 2) 双方向枝a->b,b->aのうちa->bのみ(s1がnullでない)に"W"の印をつける。
# 3) それ以外の枝は"D"として削除
f <<= nm.mcal(
c='if(isnull($s{s1}),if(isnull($s{s2}),\"F\",\"D\"),\"W\")',
a="dir")
f <<= nm.mselstr(f="dir", v="D", r=True)
f <<= nm.mcal(c='if($s{dir}==\"W\",$s{s1},$s{%s})' % (self.sim[i]),
a="sim")
f <<= nm.mchgstr(f="dir:color",
c='W:%s,F:%s' % (col[i][0], col[i][1]),
A=True)
f <<= nm.msetstr(v=[self.sim[i], str(i)], a="simType,simPriority")
f <<= nm.mcut(f="simType,simPriority,node1,node2,sim,dir,color",
o=xxfriends + "/e_" + str(i))
f.run()
# node1%1,node2%0,simType,sim,dir,color
# b,a,jaccard,0.3333333333,F,8888FF
# j,c,jaccard,0.3333333333,F,8888FF
# b,d,jaccard,0.3333333333,F,8888FF
# a,e,jaccard,0.5,W,0000FF
# d,e,jaccard,0.5,W,0000FF
# rule fileの出力
if self.orFile:
mmm = nm.mcat(i=xxfriends + "/e_*").muniq(k="node1,node2")
nm.mcommon(k="node1,node2", i=xxsimgE, m=mmm, o=self.orFile).run()
# マルチ枝の単一化(W優先,パラメータ位置優先)
if self.prune:
"""
# 双方向と片方向に分割
nm.mcat(i=xxfriends+"/e_*").mselstr(f="dir",v="W",o=xxw,u=xxf).run()
# 片方向のみの枝を選択
f = nm.mcommon(k="node1,node2",K="node1,node2",r=True,m=xxw,i=xxf)
f <<= nm.mcommon(k="node1,node2",K="node2,node1",r=True,m=xxw,o=xxff)
f.run()
f = nm.mcat(i=xxw+","+xxff).mbest(k="node1,node2",s="dir%r,simPriority%n",o=self.oeFile).run()
"""
#これだめ
fo = nm.mcat(i=xxfriends + "/e_*").mselstr(f="dir", v="W")
fu = fo.direction("u") # これは再考
fu <<= nm.mcommon(k="node1,node2", K="node1,node2", r=True, m=fo)
fu <<= nm.mcommon(k="node1,node2", K="node2,node1", r=True, m=fo)
#f = nm.m2cat()
f = nm.mbest(i=[fo, fu],
k="node1,node2",
s="dir%r,simPriority%n",
o=self.oeFile)
f.run()
else:
nm.mcat(i=xxfriends + "/e_*", o=self.oeFile).run()
nm.mcat(i=xxfriends + "/n_0", o=self.onFile).run()
def __mkTree(iFile, oFile):
temp = mtemp.Mtemp()
xxbase0 = temp.file()
xxbase1 = temp.file()
xxiFile2 = temp.file()
xxcheck = temp.file()
"""
# #{iFile}
# key,nam%0,keyNum,num,nv,nc
# #2_1,#1_1,4,1,6,1
# #2_1,#1_2,4,2,0.9999999996,1
"""
# keyNumとnum項目のuniqリストを作り、お互いの包含関係でrootノードとleafノードを識別する。
f0 = nm.mcut(f="keyNum,num", i=iFile) #{xxiFile1}
fk = f0.mcut(f="keyNum").muniq(k="keyNum") #{xxkey}
fn = f0.mcut(f="num").muniq(k="num") #{xxnum}
# root nodesの選択
fr = nm.mcommon(k="keyNum", K="num", m=fn, i=fk,
r=True).mcut(f="keyNum:node0", o=xxbase0) #{xxbase[0]}
# leaf nodesの選択
fl = nm.mcommon(k="num", K="keyNum", m=fk, i=fn,
r=True).mcut(f="num") #{xxleaf}
# leaf nodeの構造を知る必要はないので入力ファイルのnodeからleafを除外
f = nm.mcommon(k="num", m=fl, r=True, i=f0, o=xxiFile2)
nm.runs([f, fr])
def _xnjoin(inf, outf, mfile, check, no):
f = nm.mnjoin(k="node%d" % (no),
K="keyNum",
m=mfile,
n=True,
f="num:node%d" % (no + 1),
i=inf,
o=outf)
fc = nm.mdelnull(i=f, f="node%d" % (no + 1), o=check)
return fc
i = 0
depth = None
inf = xxbase0
outf = xxbase1
'''
# root nodesファイルから親子関係noodeを次々にjoinしていく
# xxbase0 : root nodes
# node0%0
# 3
# 4
# xxbase1
# node0%0,node1
# 3,
# 4,1
# 4,2
# xxbase2
# node0,node1%0,node2
# 3,,
# 4,1,
# 4,2,
# join項目(node2)の非null項目が0件で終了
'''
while True:
_xnjoin(inf, outf, xxiFile2, xxcheck, i).run()
size = mrecount(i=xxcheck)
if size == 0:
nm.msortf(f="*", i=outf, o=oFile).run()
depth = i + 1
break
# swap f_name
xxtmp = outf
outf = inf
inf = xxtmp
i += 1
return depth
def __mkNode(key, nf, nl, nv, nc, ni, ef1, ef2, ei, noiso, norm, mapFile,
oFile):
xbyE = None
xbyN = None
# edgeファイルからnode情報を生成
# noiso(孤立node排除)の場合は、edgeにあってnodeにないidを省く必要があるので計算する。
if ni == None or (ni != None and noiso):
inp = [
nm.mcut(f="%s:key,%s:nam,%s:nl" % (key, ef1, ef1), i=ei),
nm.mcut(f="%s:key,%s:nam,%s:nl" % (key, ef2, ef2), i=ei)
]
xbyE <<= nm.mnullto(i=inp, f="key", v="##NULL##")
xbyE <<= nm.muniq(k="key,nam")
xbyE <<= nm.mjoin(k="key", K="nam", m=mapFile, f="num:keyNum")
xbyE <<= nm.mjoin(k="nam", K="nam", m=mapFile, f="num,leaf")
xbyE <<= nm.msetstr(v=",,,,", a="nv,nc,nlKey,nvKey,ncKey")
xbyE <<= nm.mcut(
f="key,nam,keyNum,num,nl,nv,nc,leaf,nvKey,ncKey,nlKey")
# nodeファイルから作成
if ni:
# mcal cat用のlabel項目の作成
label = []
#label項目
if nl:
for nml in nl:
label.append(nml)
else:
label.append("$s{%s}" % (nf))
nvcStr = []
if nv:
nvcStr.append('%s:nv' % (nv))
if nc:
nvcStr.append('%s:nc' % (nc))
"""
# map
# nam,leaf,num
# ##NULL##,,0
# #1_1,,2
# #1_2,,3
# #1_3,,4
# #2_1,,5
# a,1,6
# b,1,7
# c,1,8
"""
f = None
f <<= nm.mcal(c='cat("_",%s)' % (','.join(label)), a="##label", i=ni)
if len(nvcStr) == 0:
f <<= nm.mcut(f='%s:key,%s:nam,##label:nl' % (key, nf))
else:
f <<= nm.mcut(f='%s:key,%s:nam,##label:nl,%s' %
(key, nf, ','.join(nvcStr)))
f <<= nm.mnullto(f="key", v="##NULL##")
if not nv:
f <<= nm.msetstr(v="", a="nv")
if not nc:
f <<= nm.msetstr(v="", a="nc")
f <<= nm.mjoin(k="key", K="nam", m=mapFile, f="num:keyNum")
f <<= nm.mjoin(k="nam", K="nam", m=mapFile, f="num,leaf")
if norm:
#ノードの拡大率は固定
nr = 3
f <<= nm.mnormalize(f="nv:nv2", c="range")
f <<= nm.mcal(c='${nv2}*(%s-1)+1' % (nr), a="nvv")
f <<= nm.mcut(f="key,nam,keyNum,num,nl,nvv:nv,nc,leaf") #o=#{xxa}"
else:
f <<= nm.mcut(f="key,nam,keyNum,num,nl,nv,nc,leaf") #o=#{xxa}"
xbyN <<= nm.mjoin(k="keyNum",
K="num",
m=f,
f="nl:nlk,nv:nvKey,nc:ncKey",
n=True,
i=f)
xbyN <<= nm.mcal(c='if(isnull($s{nlk}),$s{key},$s{nlk})', a='nlKey')
xbyN <<= nm.mcut(f="nlk", r=True)
if ni != None and noiso:
nm.mcommon(k="key,nam", m=xbyE, i=xbyN, o=oFile).run()
elif ni != None:
xbyN.writecsv(o=oFile).run()
else:
xbyE.writecsv(o=oFile).run()
def enumerate(self,eArgs):
"""
eArgsで与えられた条件で、頻出アイテム集合の列挙を実行する。
:type eArgs: dict
:type eArgs['type']: str
:type eArgs['minCnt']: int
:type eArgs['minSup']: float
:type eArgs['maxCnt']: int
:type eArgs['maxSup']: float
:type eArgs['minLen']: int
:type eArgs['maxLen']: int
:type eArgs['top']: int
:type eArgs['skipTP']: bool【default:False】
:param eArgs: 各種列挙パラメータ
:param eArgs['type']: 抽出するアイテム集合の型【'F':頻出集合, 'C':飽和集合, 'M':極大集合】
:param eArgs['minCnt']: 最小サポート(件数)
:param eArgs['minSup']: 最小サポート(確率)
:param eArgs['maxCnt']: 最大サポート(件数)
:param eArgs['maxSup']: 最大サポート(確率)
:param eArgs['minLen']: アイテム集合の最小アイテム数(件数)
:param eArgs['maxLen']: アイテム集合の最大アイテム数(件数)
:param eArgs['top']: 列挙するサポート上位件数(件数)
:param eArgs['skipTP']: トランザクションにマッチするパターン(アイテム集合)の出力を行わない。
"""
tf=mtemp.Mtemp()
self.eArgs=eArgs
self.type = eArgs["type"]
if "minCnt" in eArgs and eArgs["minCnt"] != None:
self.minCnt = int(eArgs["minCnt"])
self.minSup = float(self.minCnt) / float(self.db.traSize)
else:
self.minSup = float(eArgs["minSup"])
self.minCnt = int(self.minSup * float(self.db.traSize) + 0.99)
# 最大サポートと最大サポート件数
self.maxCnt=None
if ("maxCnt" in eArgs and eArgs["maxCnt"]!= None) or ( "maxSup" in eArgs and eArgs["maxSup"]!= None):
if "maxCnt" in eArgs and eArgs["maxCnt"]!= None:
self.maxCnt = int(eArgs["maxCnt"])
self.maxSup = float(self.maxCnt) / float(self.db.traSize)
else:
self.maxSup = float(eArgs["maxSup"])
self.maxCnt = int(self.maxSup * float(self.db.traSize) + 0.99)
params = {}
if self.msgoff:
params["type"] ="%sIf_"%(self.type)
else:
params["type"] ="%sIf"%(self.type)
if self.maxCnt :
params["U"] = str(self.maxCnt)
if "minLen" in eArgs and eArgs["minLen"] != None :
params["l"] = str(eArgs['minLen'])
if "maxLen" in eArgs and eArgs["maxLen"] != None :
params["u"] = str(eArgs['maxLen'])
# 列挙パターン数上限が指定されれば、一度lcmを実行して最小サポートを得る
if "top" in eArgs and eArgs["top"] != None :
self.top = eArgs["top"]
if self.top and self.top>0 :
xxtop = tf.file()
import copy
top_params = copy.deepcopy(params)
top_params["i"] = self.file
top_params["sup"] = "1"
top_params["K"] = str(self.top)
top_params["so"] = xxtop
import re
top_params["type"] = re.sub('_$', '', top_params["type"] )
extTake.lcm(top_params)
with open(xxtop, "r") as rfile:
self.minCnt = int(rfile.read().strip())
if self.minCnt<0 :
self.minCnt=1
self.skipTP=False
if "skipTP" in eArgs:
self.skipTP=eArgs["skipTP"]
# lcm_seq出力ファイル
lcmout = tf.file()
# 頻出パターンがなかった場合、lcm出力ファイルが生成されないので
# そのときのために空ファイルを生成しておいく。
with open(lcmout, "w") as efile:
pass
# lcm実行
params["i"] = self.file
params["sup"] = str(self.minCnt)
params["o"] = lcmout
extTake.lcm(params)
# caliculate one itemset for lift value
xxone= tf.file()
tpstr = "FIf_" if self.msgoff else "FIf"
extTake.lcm(type=tpstr,i=self.file,sup=1,o=xxone,l=1,u=1)
# パターンのサポートを計算しCSV出力する
#MCMD::msgLog("output patterns to CSV file ...")
xxp0 = tf.file()
self.pFile = self.temp.file()
items=self.db.items
trans0 = self.temp.file()
extTake.lcmtrans(lcmout,"p",trans0)
f = nm.mdelnull(i=trans0,f="pattern")
f <<= nm.mvreplace(vf="pattern",m=items.file,K=items.idFN,f=items.itemFN)
f <<= nm.msetstr(v=self.db.traSize,a="total")
f <<= nm.mcal(c='${count}/${total}',a="support")
f <<= nm.mcut(f="pid,pattern,size,count,total,support")
f <<= nm.mvsort(vf="pattern")
f <<= nm.msortf(f="pid",o=xxp0)
f.run()
# xxp0
# pid,count,total,support,pattern
# 0,13,13,1,A
# 4,6,13,0.4615384615,A B
xxp1=tf.file()
# taxonomy指定がない場合(2010/11/20追加)
if items.taxonomy==None:
shutil.move(xxp0,xxp1)
# taxonomy指定がある場合
else:
#MCMD::msgLog("reducing redundant rules in terms of taxonomy ...")
zdd=VSOP.constant(0)
fobj = nm.mcut(i=xxp0,f='pattern')
for fldVal in fobj:
zdd=zdd+VSOP.itemset(fldVal[0])
zdd=self.reduceTaxo(zdd,self.db.items)
xxz1=tf.file()
xxz2=tf.file()
zdd.csvout(xxz1)
f0=None
f0 <<= nm.mcut(nfni=True,f="1:pattern",i=xxz1)
f0 <<= nm.mvsort(vf="pattern")
f0 <<= nm.msortf(f="pattern")
f=None
f <<= nm.msortf(f="pattern",i=xxp0)
f <<= nm.mcommon(k="pattern",m=f0)
f <<= nm.msortf(f="pid",o=xxp1)
f.run()
# lift値の計算
transl = tf.file()
extTake.lcmtrans(xxone,"p",transl)
xxp2 = nm.mdelnull(i=transl,f="pattern")
xxp2 <<= nm.mvreplace(vf="pattern",m=items.file,K=items.idFN,f=items.itemFN)
xxp2 <<= nm.msortf(f="pattern")
xxp3 = nm.mcut(f="pid,pattern",i=xxp1)
xxp3 <<= nm.mtra(f="pattern",r=True)
xxp3 <<= nm.mjoin(k="pattern",m=xxp2,f="count:c1")
xxp3 <<= nm.mcal(c='ln(${c1})',a="c1ln")
xxp3 <<= nm.msum(k="pid",f="c1ln")
# p3
# pid,pattern,c1,c1ln
# 0,A,13,2.564949357
# 1,E,7,1.945910149
#おかしくなる?=>OK
f3 = nm.mjoin(k="pid",f="c1ln",i=xxp1,m=xxp3)
f3 <<= nm.mcal(c='round(exp(ln(${count})-${c1ln}+(${size}-1)*ln(${total})),0.0001)',a="lift")
f3 <<= nm.mcut(f="pid,size,count,total,support,lift,pattern")
f3 <<= nm.msortf(f="support%nr",o=self.pFile)
f3.run()
#self.size = mrecount.mrecount(i=self.file)
#MCMD::msgLog("the number of patterns enumerated is #{@size}")
if not self.skipTP:
# トランザクション毎に出現するシーケンスを書き出す
#MCMD::msgLog("output tid-patterns ...")
self.tFile = self.temp.file()
xxw3i = tf.file()
extTake.lcmtrans(lcmout,"t",xxw3i)
xxw1 = nm.mcut(f=self.db.idFN,i=self.db.file).muniq(k=self.db.idFN).mnumber(S=0,a="__tid",q=True).msortf(f="__tid")
xxw2 = nm.mcut(f="pid",i=self.pFile)
xxw3 = nm.mcommon(k="pid",i=xxw3i,m=xxw2).mjoin(k="__tid",m=xxw1,f=self.db.idFN).mcut(f=self.db.idFN+",pid",o=self.tFile)
xxw3.run()
def run(self):
wf = mtemp.Mtemp()
xxpal = wf.file()
xxa = wf.file()
xxb = wf.file()
xxc = wf.file()
xxd = wf.file()
xxout = wf.file()
# ============
# n1,n2,sim
# a,b,0.40
# a,c,0.31
# a,d,0.22
# b,c,0.20
# b,d,0.24
# b,e,0.14
# c,d,0.30
# d,e,0.09
xpal = None
if self.directed:
# 任意の枝a->bのaについて上位rankを選択
xpal <<= nm.mnumber(k=self.ef1,
s=self.sim + "%nr",
e="skip",
S=1,
a="##rank",
i=self.ei)
xpal <<= nm.mselnum(f="##rank", c="[," + str(self.rank) + "]")
else:
xxa = nm.mfsort(f=[self.ef1, self.ef2], i=self.ei)
xxb = nm.mfsort(f=[self.ef2, self.ef1], i=self.ei)
xpal <<= nm.muniq(k=[self.ef1, self.ef2], i=[xxa, xxb])
xpal <<= nm.mnumber(k=self.ef1,
s=self.sim + "%nr",
e="skip",
S=1,
a="##rank")
xpal <<= nm.mselnum(f="##rank", c="[," + str(self.rank) + "]")
# 両方向+片方向
xpal1 = None
if self.dir == "x":
xpal1 <<= nm.mcut(f=[self.ef1, self.ef2, self.sim], i=xpal)
# 両方向
elif self.dir == "b":
selpara = "$s{%s}==$s{##ef2}" % (self.ef1)
# 得られた上位rankグラフからa->b->cを作成し、a==cであれば相思相愛ということ
xpal1 <<= nm.mnjoin(k=self.ef2,
K=self.ef1,
m=xpal,
f=self.ef2 + ":##ef2," + self.sim + ":sim2",
i=xpal)
xpal1 <<= nm.msel(c=selpara)
xpal1 <<= nm.mcut(f=[self.ef1, self.ef2, self.sim])
else:
selpara = "$s{%s}==$s{##ef2}" % (self.ef1)
xxc = None
xxc <<= nm.mnjoin(k=self.ef2,
K=self.ef1,
m=xpal,
f=self.ef2 + ":##ef2," + self.sim + ":sim2",
i=xpal)
xxc <<= nm.msel(c=selpara)
xxc <<= nm.mcut(f=[self.ef1, self.ef2])
xpal1 <<= nm.mcut(f=[self.ef1, self.ef2, self.sim], i=xpal)
xpal1 <<= nm.mcommon(k=self.ef1 + "," + self.ef2, m=xxc, r=True)
runpal = None
kpara = "%s,%s" % (self.ef1, self.ef2)
if self.udout:
runpal <<= nm.mfsort(f=kpara, i=xpal1)
runpal <<= nm.mavg(k=kpara, f=self.sim)
runpal <<= nm.msortf(f=kpara, o=self.eo)
else:
runpal <<= nm.msortf(f=kpara, i=xpal1, o=self.eo)
runpal.run()
if self.ni and self.no:
shutil.copyfile(self.ni, self.no)